OBJECT DETECTION MODEL TRAFFIC LIGHT SYSTEM

An Undergraduate Thesis

Presented to the

Faculty of the College of Engineering

University of Rizal System

Morong, Rizal

In Partial Fulfillment

of the Requirements for the Degree

Bachelor of Science in Computer Engineering

REYDENTOR L. CASALJAY
VICTORIO DANIEL SD. CRUZ
PETER LUIS P. FENIQUITO
ERON RESTY V. SIOCO

February 2024

i
 APPROVAL SHEET

This undergraduate thesis entitled OBJECT DETECTION MODEL

TRAFFIC LIGHT SYSTEM, prepared and submitted by Reydentor L. Casaljay,

Victorio Daniel SD. Cruz, Peter Luis P. Feniquito, and Eron Resty V. Sioco, in partial

fulfillment of the requirements for the degree Bachelor of Science in Computer

Engineering is hereby recommended for approval.

February 29, 2024 PAUL ARVY V. ALFONSO, CpE

Adviser

Approved in partial fulfillment of the requirements for the degree Bachelor of

Science in Computer Engineering by the Oral Examination Committee.

CRYSTALEENE JADE A. SANTOS, CpE JULIUS CARL T. BERINGUEL, Ph.D.

Member Member

ALLAN P. ANORICO, MSEE

Chairperson

Accepted in partial fulfillment of the requirements for the degree Bachelor

of Science in Computer Engineering.

MICHAEL L. PASCUA, REE, MSCM

Date Dean, College of Engineering

ii
 ACKNOWLEDGEMENTS

The researchers express their deepest gratitude and sincere appreciation

to the following people, whose valuable help gave inspiration to them in conducting

this research study:

Engr. MICHAEL L. PASCUA, Dean of the College of Engineering, for the

allowing us to conduct the study;

Engr. FREDELINA F. DE LEON, research instructor, for the guidance and

manuals on conducting research and providing more feedback as the project

progresses;

Engr. ALLAN P. ANORICO, panel chairman, for giving suggestion,

criticism, and encouraging words for the success of the study;

Engr. PAUL ARVY V. ALFONSO, their research adviser, for the time to help

on every step of the study by providing her comments and analysis on the study;

Engr. CRYSTALEENE JADE A. SANTOS, their critic reader, for sharing

valuable comments and patiently editing this study;

Dr. JULIUS CARL T. BERINGUEL, their research expert, for contributing

his professional expertise and being open to ideas for how to improve the study;

To our beloved family, we are very grateful for your constant love and

support. We thank you for being the foundation of our lives.

And above all, to ALMIGHTY GOD, for the blessings, knowledge, spiritual

support, and eternal guidance showered to the researchers for making this

research possible.

The Researchers

iii
 DEDICATION

This dissertation is whole heartedly dedicated to the following

persons:

To our panel who shared their knowledge and helped us on

making this study possible.

To our friends who encourage, abet, and share their knowledge

to accomplish the study.

To our family for their support morally and financially to fulfill

this research.

And most of all to our Almighty God for giving us knowledge,

wisdom and courage all throughout this study.

REYDENTOR L. CASALJAY
VICTORIO DANIEL SD. CRUZ
PETER LUIS P. FENIQUITO
ERON RESTY V. SIOCO

iv
 ABSTRACT

TITLE : OBJECT DETECTION MODEL TRAFFIC

LIGHT SYSTEM

AUTHORS : Reydentor L. Casaljay

Victorio Daniel SD. Cruz
Peter Luis P. Feniquito
Eron Resty V. Sioco

COURSE : Bachelor of Science in Computer Engineering

ACADEMIC YEAR : 2023-2024

TYPE OF DOCUMENT : Undergraduate

TOTAL NO. OF PAGES : 111

NAME AND ADDRESS University of Rizal System

OF INSTITUTION : Morong, Rizal

EXECUTIVE SUMMARY:

This research study entitled Object Detection Model Traffic Light System

was conducted during the School Year 2023 - 2024 in Pahati Bridge at Barangay

Dalig, Teresa, Rizal. The research general objective of this research is to evaluate

the developed Object Detection Model Traffic Light System.

To determine the performance efficiency and level of acceptability of the

study, the researchers used Intersection over Union, and mean. On the other hand,

the standard deviation and weighted mean were used to determine the level of

acceptability of the system.

The researchers utilized a combination of developmental and descriptive

methods of study. The number of respondents in this research is determined by

systematic sampling. Additionally, researchers utilized a questionnaire checklist to

gather data from respondents. This questionnaire checklist is designed based on

v
ISO/IEC 25010 which includes functional suitability, performance efficiency,

compatibility, reliability, usability security, and maintainability. The data gathered

from the questionnaire is interpreted using a Likert scale, and ODM efficiency

table.

The computed mean of the object detection model efficiency is 0.8603 with

an interpretation of “Very Much Efficient”. Moreover, the propagation delay resulted

to mean of 0.32420 seconds.

On the other hand, the results of the evaluation showed that the developed

device and system was highly acceptable, with an average weighted means of

4.58, 4.55, 4.53, 4.59, 4.40, 4.56 and 4.54 from the Community which is the

drivers, experts and the local government unit respectively. These findings were

interpreted as "Very Much Acceptable".

In this study, the researchers found that the integration of the Video

Capturing Device (RPi Camera V2), Object Detection Model (Yolo V5 Nano 6),

Logic Devices (Raspberry Pi 4 and Node MCU ESP8266), Wireless

Communication (Comfast EW73), and Traffic Signal Light (LED) successfully

performed its specified functions, getting the highest score in its usability.

The researcher’s recommendation to future researchers is to make the

dataset more accurate, use a higher specification Raspberry Pi, a wider range and

good signal quality, and use a renewable power source. Future researchers may

consider integrating additional features into the system’s hardware.

vi
 TABLE OF CONTENTS

Page

TITLE PAGE……………………………………………………………………. i

APPROVAL SHEET…………………………………………………………… ii

ACKNOWLEDGEMENTS……………………………………………………. iii

DEDICATION………………………………………………………………….. iv

ABSTRACT……………………………………………………………………. v

TABLE OF CONTENTS………………………………………………………. vi

LIST OF TABLES ..……………………………………………………….….… viii

LIST OF FIGURES……………………………………………………….……. ix

Chapter

1 BACKGROUND OF PROJECT

Background of the Study……………………………………. 1

Objectives of the Study……………………………………… 11
Theoretical Framework……………………………………… 12
Conceptual Framework……………………………………… 13
Significance of the Study……………………………………. 16
Scope and Limitations……………………………………….. 17
Definition of Terms…………………………………………… 19

2 DESIGN METHODOLOGY

Research Technical Design………………………………… 22

Research Instruments………………………………………. 32
Procedure and Method of the Study………………………. 35
Project Design Models……………………………………… 37
Statistical Treatment…………………………………………. 40

3 PRESENTATION, ANALYSIS, AND DESIGN RESULTS

Design and Development of an Object Detection Model

Traffic Light System…………………………………………. 42
Test the Performance Efficiency of the Developed Object
Detection Model Traffic Light System………...……………. 52

vii
 Level of Acceptability of the Developed Object
Detection Model Traffic Light System……………………. 55

4 SUMMARY OF DESIGN RESULTS, CONCLUSIONS,

AND RECOMMENDATIONS

Summary of Design Results………………………………. 67

Conclusions……………………………………………….… 68
Recommendations…………………………………………… 69

REFERENCES……………………………………………………………… 71

APPENDIX

A Gantt Chart of Activities……………………………………. 75

B Questionnaire Checklist………………………………….... 78
C Certificate of Content Validation………………………… 83
D Letter to Conduct The Study………………………………. 84
E Letter for the Respondents………………………………… 85
F Letter to the Panel Members………………………………. 86
G Statistical Formula Used……………………………………. 90
H Summary of Components and Equipment Used…….…. 94
I List and Cost of Materials………………………………….. 99
J User’s Manual………………………………………………. 102
M Random Pictures Taken During the Conduct of
Evaluation……………………………………………….…... 104

CURRICULUM VITAE……………………………………………………… 107

viii
 LIST OF TABLES

Table Page

1 Computed Mean of Object Detection Model Efficiency of the

developed Object Detection Model Traffic Light System……... 52

2 Computed Mean of Propagation Delay of the developed

Object Detection Model Traffic Light System ….……………… 54

3 Computed Weighted Mean on the Level of Acceptability of the

developed Object Detection Model Traffic Light System in
Terms of Functional Suitability …………………………………. 56

4 Computed Weighted Mean on the Level of Acceptability of the

developed Object Detection Model Traffic Light System in
Terms of Compatibility…………………………………………… 57

5 Computed Weighted Mean on the Level of Acceptability of the

developed Object Detection Model Traffic Light System in
Terms of Reliability………………………………………………. 59

6 Computed Weighted Mean on the Level of Acceptability of the

developed Object Detection Model Traffic Light System in
Terms of Usability………………………………………………… 60

7 Computed Weighted Mean on the Level of Acceptability of the

developed Object Detection Model Traffic Light System in
terms of Security……………………………………………….… 62

8 Computed Weighted Mean on the Level of Acceptability of the

developed Object Detection Model Traffic Light System in
terms of Maintainability …………………………………………. 63

9 Composite Table of the Level of Acceptability of the

developed Object Detection Model Traffic Light System……... 65

ix
 LIST OF FIGURES

Figure Page

1 Conceptual Model of Object Detection Model Traffic Light

System……………….…………………………………………… 15
2 Block Diagram of YOLO Object Detection Model
Training.............................................................................. 23

3 Block Diagram of Computer Vision Feature…………………… 24

4 Object Detection Model………………………………………….. 26

5 Dataflow Diagram of Object Detection Model Traffic Light

System…………………………………………………………..... 27

6 Block Diagram of Object Detection Model Traffic Light

System……………………………………………………………. 29

7 Flow Chart of Object Detection Model Traffic Light System….. 31

8 RPi Camera V2 (Computer Vision)....………………………….. 43

9 Yolo V5 N6 (Object Detection Model) .…………………………. 44

10 Raspberry Pi Microcontroller (Logic Device 1)………………... 46

11 Comfast EW73 Wifi Access Point………………………………. 47

12 Node MCU ESP8266 (Logic Device 2)………………...………. 49

13 Light Emitting Diode (Traffic Signal Light)…..…………………. 50

x
 1

Chapter 1

BACKGROUND OF PROJECT

This chapter deals with the background of the study, objectives of the study,

theoretical framework and conceptual framework, significance of the study, scope

and limitations, and the definition of terms.

Background of the Study

Technology is defined as the application of knowledge to achieve a goal. It

is also defined as a tool for people to make tasks easier and more efficient. Many

people think that technology is only present in modern times. But the truth is,

technology was already present in the Stone Age. These technologies are the

stone tools that the human ancestors used. These stone tools help humans not

only to survive but also to make life better. The technology evolves from stone to

metal. Metals are still used today because of their qualities. These metal tools only

change in shapes, design, features, and application. It is also the same with

modern technology. It only evolves based on human needs. These advancements

are based on the aim of technology, which is to make human life better.

Artificial Intelligence is a technology that places the human intelligence in a

machine. Specifically, it is a simulation of human intelligence in a machine.

According to Frankenfield (2023), artificial intelligence, or AI, refers to the

simulation of human intelligence by software-coded heuristics. Nowadays this

code is prevalent in everything from cloud-based, enterprise applications to

consumer apps and even embedded firmware. In this study, the researchers

utilized the artificial intelligence. The artificial intelligence served as a vision of the
 2

device. Through artificial intelligence, the machine can recognize objects and

perform logical operations.

Computer vision, as its name implies, is a vision of the computer. Computer

vision is also known as a field of artificial intelligence. It is known to have a

numerous algorithms and data in order to accurately recognize an object.

According to Ashtari (2023), computer vision leverages artificial intelligence (AI) to

allow computers to obtain meaningful data from visual inputs such as photos and

videos. The insights gained from computer vision are then used to take automated

actions. Just like AI gives computers the ability to ‘think’, computer vision allows

them to 'see'. This article defines computer vision, which is one of the many

features of the device in this study.

In the study conducted by Delgado et al., (2023), they developed a device

that identifies vehicles in a parking lot using optical camera recognition (OCR),

which is artificial intelligence. Their research has features like reservations, a car

parking data sheet, a display for parking slots, and plate recognition. Both

Delgado's research and this research used artificial intelligence and cameras

which falls under computer vision.

According to an article written by Umali (2019), there are students who have

developed software that would help authorities quickly and more efficiently identify

traffic violators. CATCH-ALL, or the Contactless Apprehension of Traffic Violators

on a 24-Hour Basis and All-Vehicle Detection System, is capable of detecting and

tracking vehicles. The closed-circuit television (CCTV) cameras installed are

equipped with artificial intelligence (AI) software for real-time monitoring of traffic
 3

situations. In this study, the researchers also developed a device equipped with a

camera and artificial intelligence. Specifically, the field of artificial intelligence is

computer vision.

Furthermore, based on the study written by Ecaldre et al., (2022), using pre-

trained deep learning and a color-connected component algorithm, the prototype

model was successfully simulated using an image processing tool. The prototype

was able to manage traffic at a particular flow, giving priority to the green light

where there were more cards in one lane than the other. In their study, they

concluded that using pre-trained deep learning, they could lessen congestion and

waste time at traffic lights. But there are limitations to their device, like accuracy.

The accuracy of their device is dependent on the quality of the camera. The

researchers also noticed that the object identification algorithm used in their device

is based on color, which is not desirable for road implementation. The limitations

mentioned is solved in this research. Hence, 1987 Constitution of the Republic of

the Philippines, Article 14, Section 10 and 12 states that:

“Science and technology are essential for national development

and progress. The State shall give priority to research and
development, invention, innovation, and their utilization; and to
science and technology education, training, and services. It shall
support indigenous, appropriate, and self-reliant scientific and
technological capabilities and their application to the country's
productive systems and national life.
The State shall regulate the transfer and promote the adaptation
of technology from all sources for the national benefit. It shall
encourage the widest participation of private groups, local
governments, and community-based organizations in genration
and utilization of science and technology.”

The cited provision is a series of sections that pertain to the national

develoment of the country through education, science and technology, arts,

 4

culture, and sports. The researchers conducted this research in response to

Section 10 and Section 12.

In this study, the researchers aim to develop a device that detects a specific

object, which is why they chose to use the object detection model of artificial

intelligence. The object detection model that the researchers will use is YOLO, or

You Only Live Once, by Ultralytics. According to Kundu (2023), object detection

deals with localizing a region of interest within an image and classifying this region

like a typical image classifier. One image can include several regions of interest

pointing to different objects. This makes object detection a more advanced

problem of image classification. YOLO (You Only Look Once) is a popular object

detection model known for its speed and accuracy. It was first introduced by

Joseph Redmon and others in 2016 and has since undergone several iterations,

the latest being Yolo v7. YOLO is the object detection model that was utilized in

this research because of the aforementioned characteristics.

Running software requires hardware. Hence, the logic device is needed to

store and execute the program. The primary logic device in this study is the

Raspberry Pi. According to an article written by Valerio (2023), Arm Holdings plc

has announced that they have made a strategic investment in Raspberry Pi. The

Raspberry Pi was originally intended for students, but it has become a tool to

develop Internet of Things (IoT) prototypes and items. After securing their

agreement with ARM, the latter company gained a minority stake in the Raspberry

Pi. It is said in the article that IoT solutions will be more accessible to students and

anyone by lowering the barriers to innovation. ARM is known for making

 5

smartphone chips and single-board computers. With this move, the Raspberry Pi

will make the single-board computer market affordable and within the reach of

anyone. This relates to the study because the device in this study utilized a

Raspberry Pi for harboring artificial intelligence, and the reason for choosing this

product is because of its processing power and affordability.

The Raspberry Pi 4b is the chosen main logic device that harbors a deep

learning model. It was chosen for its low power consumption, affordability, and

processing power. In the next article, the authors conducted a comparison of

performance between three logic devices in executing Yolo V3. According to Feng

et al., (2023), they found out that the Jetson Xavier outperformed the Raspberry Pi

4B and Jetson Nano in running Yolo V3. Also, they noticed that the highest

performing device also has the highest power consumption.

The secondary logic device that will be used in this research is the

NodeMCU ESP 8266. NodeMCU ESP8266 is the same as Arduino. They are both

programmable. The difference is only in the wireless capability of both

microcontrollers. This wireless capability is commonly used in the Internet of

Things, or IoT. According to the journal written by Yasay (2021), the Internet of

Things (IoT) is a global knowledge society system that enables advanced services

by interconnecting (physical and virtual) things using current and emerging

interoperable information and communication technologies. Because of the

convergence of multiple technologies, such as real-time analytics, machine

learning, commodity sensors, and embedded systems, things have changed. In

his journal, he also mentioned that dumb devices can be smartened because of
 6

these IoT solutions that are available on the market. This explains the reason for

the researchers' choice of logic device. It needs to be capable of communicating

wirelessly. Specifically, the researchers are avoiding bulky and unorganized wired

connections between two logic devices.

Moreover, Node MCU ESP8266 is also responsible for the storage of the

password and username of the power control web-panel. Now, to assess the level

of security of the system, here is an article about securing authentication data. In

an article written by Blue et al., (2017), there are numerous ways to protect

database storage that stores passwords. They mention the hashing, where the

passwords are hashed to protect them even if attackers have access to the

database. Another security measure is to provide a password policy that enforces

length, entropy, and lifespan. This may provide safe guards against password

guessing or brute-force attacks. In the system developed in this research, the web-

panel has a function to change passwords. This is the security measure that the

researchers implemented to prevent brute-force attacks. Additionally, the security

of the system in this research is protected by the wireless network by preventing

unauthorized IP addresses and Mac addresses from accessing the network.

The wireless technology that will be utilized in this research is Wi-Fi. To

make this happen, the researchers planned to have Wi-Fi that is capable of

communication over a wide distance. For this reason, the researchers decided to

use an outdoor Wi-Fi AP. In particular, it is the Comfast EF-EW73. When using

wireless technology, the users must be knowledgeable about its disadvantages.

Hence, according to an article written by Sonawane et al., (2017), the very slow
 7

speed of the network leads to limited connectivity and delays in processing hours,

so people tend to use wireless technology either at home, at a coffee shop, or at

the airport. People deal with the different ways of wireless networking. Wi-Fi is the

most accepted technology and is the wireless way to handle networking. This

networking might be interrupted either by server failure or by a loss of range. They

also mentioned that this lost connection is caused by Wi-Fi dead zones. To prevent

this, the Wi-Fi owner should place their Wi-Fi device in strategic zones. Also, they

can add or replace devices that can help widen and strengthen the signal. The

researchers considered the data and conclusion of their study in planning the

wireless connectivity of the two logic devices in this research.

The aforementioned features are the list of feature objective of the device

that will be developed in this research. However, the purpose of the device must

be stated. Traffic congestion is the problem that the device aims to solve. . In the

Philippines, traffic is a big problem in transportation. There are several factors that

contribute to this: one is the lack of urban planning by the city government, and the

other is the lack of infrastructure projects that will significantly help traffic road

management. There are numerous systems on the market that are applicable to

this problem. But they are not all designed to be compatible with different locations.

For this reason, the researchers developed a traffic solution that is intelligent and

compatible with the place where the research will be implemented.

According to the study of Fattah et al., (2021), they found that excessive

vehicles, narrow roads, ineffective public transport modes, and bus operators’

attitudes are causing congestion, resulting in high to extreme stress levels for road
 8

users, with a total economic loss of $2.01 million per day. Traffic congestion is a

problem that researchers are aiming to solve. Traffic congestion is a vital problem.

It is not only affecting the country’s economy but also the community. Thus,

Republic Act No. 4136, Section 35, states that:

“Restriction as to Speed. – (a) Any person driving a motor vehicle

on a highway shall drive the same at a careful and prudent speed,
not greater nor less than is reasonable and proper, having due
regard for the traffic, the width of the highway, and of any other
condition then and there existing; and no person shall drive any
motor vehicle upon a highway at such a speed as to endanger the
life, limb and property of any person, nor at a speed greater than
will permit him to bring the vehicle to a stop within the assured
clear distance ahead."

The cited provision is a series of sections that pertains to an act to compile

the laws relative to land transportation and traffic rules, to create a land

transportation commission and for other purposes. The researchers conducted this

research in accordance to Section 35.

In the next study, the researchers develop a traffic flow prediction using

machine learning. Based on the study of Espinoza et al., (2022), neural networks

are computing systems inspired by the biological neural networks in human brains.

They are trained in a batch learning mode; hence, the whole training data should

be ready before the training task. However, this is not applicable for many real-

time applications where data arrives sequentially, such as online topic detection in

social communities, traffic flow prediction, etc. In this paper, an efficient hardware

implementation of a low-latency online neural network system is proposed for a

traffic flow prediction application. Their study relates to this study because, like

them, the device developed in this research is for the purpose of traffic. The
 9

researchers used their study as a guide in the field of hardware and artificial

intelligence in order to make this study successful.

Furthermore, there is also a study about the impact of autonomous vehicles.

According to Corpus et al., (2023), there are still some traffic symbols that have a

noticeable rate of inaccurate identification by drivers. This was caused by the

design. They also emphasized that there is room for improvement in visual design.

In their recommendation, they said that authorities could provide educational

training to increase traffic sign knowledge. On the other hand, based on their study,

autonomous vehicles proved that there is a great result in road penetration,

resulting in good traffic flow on roads. They also mentioned that autonomous

vehicles can be improved to stay up-to-date with changes in their environment. By

doing so, autonomous vehicles will become more accurate and efficient. This study

shows how artificial intelligence can be applied to road traffic. It can be gleaned

from their study that both humans and autonomous vehicles have a lot of room to

improve.

Here is another study, but this one focuses on smart utilization of electricity.

According to Widartha et al., (2023), increasing demand for energy-efficient and

sustainable indoor environments has led to the development of various smart

lighting and energy consumption management systems. In their study, they

developed a smart lighting system that can efficiently reduce electricity usage. This

was done by making a system integrated with a sensor to sense the amount of

light entering the room. Additionally, they added daylight data to accurately adjust

the lighting. Their study might not be under artificial intelligence, but it is smart
 10

enough to run automatically. In this study, the researchers made a system that can

automatically adjust the lighting in the camera for better detection of the object

detection model. This idea was used by the researchers to make the system more

efficient.

The cited literature and studies are all related to the development of this

research. Some of them are features that will be implemented in this study.

Specifically, those features are aimed at providing a better and more modern traffic

solution for the community. The big problem with the roads here in the Philippines

is that they cannot handle the volume of cars. Some of them are providing paths

in two directions, even though the size of the road is one way. These problems are

present at the research location. Road congestion is happening in that area

because there are schools and markets nearby. To solve this, the researchers

developed a solution to reduce road congestion, and that is by implementing an

Object Detection Model Traffic Light System.

The Object Detection Model Traffic Light System aimed to solve the traffic

and provide a systematic solution to the community near the Pahati Bridge.

Through this device, the traffic congestion will lessen. Using an object detection

model called YOLO, the traffic lights will be intelligent. It can recognize vehicles

and perform intricate logical operations and conditions. It is also integrated with

wireless technology, with the use of access point Comfast EW-73, to provide

communication between two traffic lights, making it compact and small. The

researchers also ensure that there are minimal delays with this device by making

the operations as real-time as possible.

 11

Objectives of the Study

The general objective of this research is to evaluate the Object Detection

Model Traffic Light System at Pahati Bridge in Barangay Dalig Teresa, Rizal.

Specifically, the study sought to answer the following;

1. To design and develop Object Detection Model Traffic Light System in

consideration of the following features:

1.1 Computer Vision (RPi Camera V2);

1.2 Object Detection Model (Yolo V5 Nano 6);

1.3 Logic Device 1 (Raspberry Pi 4);

1.4 Wireless Communication (Comfast EW73);

1.5 Logic Device 2 (Node MCU ESP8266); and

1.6 Traffic Signal Light (LED).

2. To test the performance efficiency of the developed Object Detection

Model Light Systems’ capacity having four vehicles camera range, 3220

ms processing the weights, and 90 meters Wi-Fi range through:

2.1 Object Detection Model; and

2.2 Propagation Delay.

3. To determine the level of acceptability of the developed Object Detection

Model Traffic Light System in terms of:

3.1 Functional Suitability;

3.2 Compatibility;

3.3 Reliability;

3.4 Usability;
 12

3.5 Security; and

3.6 Maintainability.

Theoretical Framework

In this study, there are two main ideas. The first one is the traffic, which is

the problem. The second is object detection, which is the solution of the

researchers to the problem. With these in mind, the researchers find two related

theories that elaborates these ideas. These two theories are the Traffic Flow

Theory by Wu (2021) and the Object Detection Theory by Li (2020). The

researchers believed that the aforementioned theories were related to this

research.

According to Wu (2021), traffic flow theory deals with models and

hypotheses for explaining traffic flow, i.e., what would happen to traffic streams if

they were to flow on roads under different conditions, potentially not yet observed.

In this study, the researchers developed a traffic signalling device that will

introduce a new traffic scheme for the road users in the research location. This

device is controlled the traffic flow at the research location, making the travel of

vehicles in both opposite directions organized. The researchers will also observe

the traffic flow to see if there are significant changes. System changes will be made

if there are problems at the research location.

Furthermore, the study was based on object detection theory. According to

Li (2020), with the continuous improvement and development of the object

detection technology, there are more ideal use experience and a wide range of

applications. In the aspect of traffic, object detection can be efficiently completed

 13

by detecting pedestrian, vehicle, road signs, traffic lights and other objects on the

road to assist traffic management. The researchers are trying to apply advanced

technologies such as ODM or Object Detection Model to speed up traffic

resolution. It is possible to monitor traffic conditions in real-time by deploying object

detection models. These models are capable of tracking and identifying cars,

citizens, and other objects by looking at footage captured from traffic cameras.

This information can be used to assess traffic flow, detect congestion, and identify

areas that may need traffic management interventions.

By developing a Traffic Light System Using Object Detection Model, the

researchers aimed to reduce road congestion and the number of road rages. This

research will greatly save time, money, and, most importantly, the lives of people.

The principles that have been mentioned before will serve as guides for the

researchers in order to achieve the objectives of this research.

Conceptual Framework

The methodology and strategy that were used to develop this study were

based on the innovation life cycle. Specifically, the Innovation Life Cycle of

Improvised was used. The researchers modified and adopted the model based on

the stages taken to develop the device in this research.

According to Larkins (2021), an “innovation lifecycle” is the different stages

of the innovation process, including the conditions and factors that support each

stage and how and when to use different tools and ways of doing things. The

innovation life cycle was modified based on the objectives of this research. Each

stage taken in the innovation life cycle will satisfy the aim of this research. This
 14

model will ensure that the researchers are following the standard phases of idea

innovation.

In the investigate stage, the researchers gathered data from the people who

experienced using the road where the research conducted. They are the tricycle

drivers, vehicle owners of Carissa 2, and people living near Pahati Bridge in

Barangay Dalig Teresa, Rizal. The researchers also inspect the environmental

condition of the research location. Based on these data, the researchers chose

materials, components, frameworks, and programming libraries.

During the analysis stage, the researchers analyzed the data that they have

gathered. These data are sorted out based on its importance and impact.

From there, they selected the feasible and applicable ones for this research.

The next stage is the design, develop, and test stage. This is the stage

where the researchers used all the modules, frameworks, components, and

programming libraries they have gathered. This is also the stage where the

researchers developed a solution for the concerns and problems they filtered from

the analysis stage.

On the other hand, in the stage labelled as level of acceptability, the solution

gets implemented. This is the stage where the researchers implemented the Object

Detection Model Traffic Light System at Pahati Bridge in Barangay Dalig Teresa,

Rizal. The written program is tested based on its expected functionality and design.

The researchers also conducted a survey to gather feedback from the

respondents. The gathered feedback is used as a guide for the researchers in

modifying and upgrading the Object Detection Model Traffic Light System.
 15

Figure 1
Conceptual Model of Object Detection Model Traffic Light System
 16

The last stage is where the Object Detection Model Traffic Light System is

fully developed. In this stage, the researchers consulted experts. After the

consultation, they will focused on maintaining the system. If the bug appears during

this stage, the researchers will support the device with updates and patches. If

there is any innovation that needs to be done, then the researchers will just repeat

the cycle again.

Significance of the Study

This research aims to design, develop, and provide a system that can lead

to the modernization of the following beneficiaries:

Community. By implementing the Object Detection Model Traffic Light System, the

study seeks to raise awareness and promote adherence to traffic safety rules

among the residents. To improve traffic management, it aims to contribute to a

decrease in traffic accidents, enhancing overall safety and well-being within the

community.

Local Government Unit. The study developed an Object Detection Model Traffic

Light System specifically designed to address the common problems of traffic

congestion, road rage, and urban challenges in Pahati Bridge Barangay, Dalig,

Teresa, Rizal. By providing real-time data and insights, this system can aid local

authorities in making informed decisions to alleviate traffic-related problems and

enhance urban mobility.

Future Researchers. This research offers significant value not only for its current

applications but also for future transportation engineering researchers and

practitioners. It produces a greater comprehension of the subject matter by

 17

thoroughly exploring the complex nature of traffic systems and possible paths for

change. Furthermore, it enables future researchers to expand on current

understanding and produce advanced solutions to traffic management problems

by creating foundations for future innovation and improvement.

Scope and Limitations

This research entitled “Object Detection Model Traffic Light System,” was

conducted by students from the College of Engineering at the University of

Morong Rizal System during the first and second semesters of the academic year

2023-2024. The researchers aim to develop a system that will modernize traffic

management.

The data that will be collected is at the location of the Barangay Dalig at

Pahati Bridge, from January to April 2024.

The target audience of this study are the Drivers, and the Pedestrian that

using the Pahati Bridge. The researchers will then give a set of questionnaires to

the target audience so that they can collect data. With the collected data, the

researchers will be able to quantitatively portray what the target audience will

experience, what their qualitative experience will be in using this study, and what

their suggestions and recommendations are for this study.

This study is focused to design and construct a prototype of a Traffic Light

System. The system's most important feature is the Raspberry Pi. Using an RPI

V2 Camera that will be placed on the other side of the bridge and an Access Point

that will cover the distance.

 18

The researchers will develop an Object Detection or YOLO on the

Raspberry Pi that will be used to identify the vehicle. As a result, traffic control

will be more practical and efficient. This approach will reduce traffic jams and

road rage. The device comes with a user manual to assist the barangay official

in using it.

The methodology and strategies that were used to conduct the research

were a combination of development research in order to properly design and

develop the system and a descriptive study in order to describe the paper

accurately and systematically.

The researchers set a 50-meter distance limit for the camera because of its

8 megapixels to prevent inaccurate results.

To minimize process overload and reduce processor workload, the YOLO

divides the tracking of the objects into three groups.

 19

Definition of Terms

For a better understanding of the study, the following terms were defined

operationally by the researchers.

Capacity. The term refers to the degree to which the system deems the detected

objects or situations as acceptable or safe.

Comfast EW73. The term refers to a wireless network adapter which is designed

to provide Wi-Fi connectivity to devices that lack built-in wireless capabilities or to

enhance the wireless performance of existing devices.

Compatibility. The term refers to the ability of software and hardware from different

sources to work together without having to be altered to do so. This means that

programs, devices and systems must interact with each other without issues

arising.

Design and Develop. The terms refers to process of creating a visual or conceptual

representation of a product or service. On the other hand, development is the

process of creating and implementing a service or product.

Functionality Suitability. The term refers to the ability of the system to recognize

and process real-time environmental data so that it can decide when to change

traffic signals.

Level of Acceptability. The term refers to the degree or extent to which something

is deemed suitable, satisfactory, or permissible within a given context. It can be

applied to various situations and fields.

Light Emitting Diode (LED).The term refers to a semiconductor device that emits

light when an electric current is applied to it.

 20

Maintainability. The term refers to the probability that, in case of failure, an

assigned technician will be able to restore or repair a specific problem within the

device, both software and hardware.

Node MCU ESP8266. The term refers to a low-cost open-source IoT platform

which combines a microcontroller with built-in Wi-Fi capabilities, making it suitable

for various IoT applications.

Object Detection Model Efficiency. The term refers to the capability of a computer

vision system to accurately and swiftly identify and locate objects within images

or video streams while optimizing various critical aspects.

Object Detection Model Traffic Light System. The term refers to a device or system

designed to identify and locate vehicles within an image or a video stream and

knowing which lane will stop and go.

Test the Performance Efficiency. The term refers to the performance relative to the

device’s ability to execute its functions consistently and optimally, particularly in

terms of accurately identifying and responding to objects within its operational

domain.

Propagation Delay. The term refers to how long it takes a signal to travel over a

network from its sender to its receiver. As computers send signals between one

another, there are delays based on the distance between the two routers.

Raspberry Pi 4. The term refers to a credit card-sized single-board computer

developed by the Raspberry Pi Foundation.

Reliability. The term refers to the ability of the system to perform its intended

functions correctly and consistently over a specified period of time and under
 21

various operating conditions.

RPi Camera V2. The term refers to which is a camera accessory designed for use

with the Raspberry Pi single-board computers.

Security. The term refers to measures and features implemented to ensure the

integrity, confidentiality, and availability of the system.

Usability. The term refers to measures a product's ability to help users achieve

their objectives.

Yolo V5 Nano 6. The term refers to "You Only Look Once," and it is a real-time

object detection algorithm. YOLO is popular for its speed and accuracy in detecting

objects in images and videos.

 22

Chapter 2

DESIGN METHODOLOGY

This chapter represents the research technical method, research

instrument, procedures and method of the study, project design model, and lastly

statistical treatment.

Research Technical Design

The methodology and strategies that is used to conduct the research are a

combination of developmental and a descriptive research design.

According to Klaassen et al., (2015), developmental research is a particular

way of addressing the basic questions of why and how to teach what to whom. The

researchers utilized the developmental methods of developmental research in

developing the device in this research. This also served as guidelines for the

researchers in generating additional ideas and developing the system. Making

sure that the system developed in this research follows standard system quality

and procedures.

On the other hand, based on McCombes (2019), the descriptive method is

defined as a method that aims to accurately and systematically describe a

population, situation, or phenomenon. The descriptive method is utilized by the

researchers in order to determine the population and variables of this research.

The researchers also used the descriptive method to determine the level of

acceptability of this research.

Furthermore, the aforementioned methods of research contribute to this

study by making it feasible and systematic. First, the developmental method

 23

specifically makes the process of developing the system convenient. One of the

developmental methods utilized in this research is the gathering of information by

the researchers. By doing this, the system in this research will become compatible

with the problem in the study location. This also makes the research aligned with

its specific objective, which is not only to provide a system that solves the problem

but also to provide a long-term solution that is efficient and modern. Additionally,

this study utilized descriptive methods of research in order to systematically

observe the variables in this research. These variables, like occupation, will greatly

affect the conclusion of the study. Any significant values in the variables gathered

in the survey will be stated in the conclusion and this serves as a systematic

presentation of the results of this research.

Figure 2
Block Diagram of YOLO Object Detection Model Training

In Figure 2, it shows the process that the researchers used to train the Yolo

Object Detection Model. In the first block, the researchers gathered a labeled

dataset of vehicles in the Philippines. It includes five classes, which are cars,

tricycles, trucks, and motors. After getting the labeled data, the researchers trained

the object detection model. In the final process, the researchers acquired the Yolo

weights that are used for object detection.

 24

A dataset is a set of picture data from a specific class. For instance, the

vehicle pictures in the dataset of this research. These datasets can be labeled or

unlabeled. The ones that the researchers used are labeled. In research, object

detection requires a labeled dataset in the training process. The output of this

process is the weights that are used in object detection.

Figure 3
Block Diagram of Computer Vision Feature

Next is Figure 3, which illustrates the computer vision process. In the first

block, it shows the camera running on 1 frame per 4 seconds. The 1 frame per 4

seconds is the expected maximum speed for video capture of data because of the

low processing power of the Raspberry Pi 4. After that is the block showing the

Raspberry Pi 4 storing and executing the Yolo object detection model.

In the first box of Figure 3, it contains the Raspberry Pi V2 camera with 1

frame per 4 seconds of expected speed. However, this low speed of capture is not

caused by the camera but by the processing. The system structure has a
 25

bottleneck because the Raspberry Pi can’t handle rendering video data while also

detecting. Based on the documentation of the Raspberry Pi V2 camera, it can

capture 203 fps with 640x480 resolution, but in this system, it was reduced to 1

frame per 4 seconds, or 0.14 FPS.

Next box is the Raspberry Pi 4 with a YOLO Object Detection Model inside.

The box 4 indicates that the logic device which is Raspberry Pi 4 is the one that

executes the object detection model. This execution utilizes its resources like

cores, and ram resulting to lower speed of detection

The overall process of the computer vision feature is simple. The task of the

first is to serve as the sensor for the system. It collects video data to feed into the

object detection model. In the other smart or intelligent traffic light, the system

either has an ultrasonic, infrared, or proximity sensor. But in this study, the

researchers utilized a camera as an input for the system. On the other hand, the

second box serves as the brain to perform logical tasks. These tasks are process

the weights and recognize the vehicles based on the weights.

In Figure 4, it shows the procedures of YOLO in identifying and detecting

the model. According to Chablani (2017), the system divides the input image into

an S × S grid. If the center of an object falls into a grid cell, that grid cell is

responsible for detecting that object. Each grid cell predicts B bounding boxes and

confidence scores for those boxes. These confidence scores reflect how confident

the model is that the box contains an object and also how accurate it thinks the

box is that it predicts. Formally they define confidence as Pr(Object) ∗ IOU . If no

object exists in that cell, the confidence scores should be zero. Otherwise they
 26

want the confidence score to equal the intersection over union (IOU) between the

predicted box and the ground truth.

Figure 4
Object Detection Model

In the first box of Figure 4, the object detection model plots the S x S grid

on the video data received. The purpose of this is to make the detection and

classification of the object detection model easier. Next process is the object

detection model, which performs a plotting of the bounding box and class

probability map. The bounding boxes are plotted based on target classes of

weights, like dogs and bikes in the picture. This plotting of bounding boxes is

random in this phase. This will be filtered later based on the classification and

confidence of detection. On the other hand, the class probability map is the phase

of detection where the object detection is coloring the S x S grid by its classification.
 27

This process will make the classification and plotting of bounding boxes accurate.

In the last box, the plotting bounding box and classification are finished and

combined, making the object detection accurate from its true classification and

ground truth bounding boxes.

Figure 5
Dataflow Diagram of Object Detection Model Traffic Light System

In Figure 5, it shows the flow of data from input to output. In the first block,

the device needs to gather video data using a camera. The next block shows the

Rasperry Pi 4, which is responsible for processing the video data coming from the

camera. After that, the Raspberry Pi 4 produces a result through traffic light output

and wireless AP. The data sent to wireless AP is now transported to the logic

device 2, which is the Node MCU ESP8266. The logic device 2 now reads the data

and executes the appropriate output based on the received data.

 28

In the first box of Figure 5, it contains a camera. Specifically, the camera

that the system utilizes is the Raspberry Pi V2 camera. The task of this component

is to gather real-world video data. The video data acquired is sent to logic device

1, which is the Raspberry Pi 4. The Raspberry Pi 4 is tasked with rendering the

video data coming from the camera. The logic device 1 is also responsible for

executing the object detection that classifies the objects in the video data. Logic

device 1 is now performing an execution of the logic condition based on the object

detection. An output is shown on the traffic light. This was shown in the bottom box

in Figure 6. The next thing that the system does is transfer the data to logic device

2. This transfer is done through wireless. The device that was used for wireless

communication is the Comfast EW73. The endpoint of wireless communication is

logic device 2. The message sent by logic device 1 is now processed by logic

device 2. Based on it, the logic device 2 shows an output by means of a traffic light.

In Figure 6 block diagram, the researchers illustrate the overview of the

components of the smart traffic light system. First is the DC power supply, where

both the Raspberry Pi and the Node MCU get their power. The two logic devices

have a separate DC power supply. The next block is the Raspberry Pi 4. This

device is responsible for executing the logic instructions of Road 1 (Callejon 3).

Next is the RPI camera, which is responsible for capturing video data for the AI

object detection model. After getting the video data, the object detection model,

which is executed by the Raspberry Pi 4, is responsible in recognizing the vehicles.

The object detection model is fed trained data before implementation of the device.

After that, the traffic lights shows a traffic signal based on the instructions of logic
 29

Figure 6
Block Diagram of Object Detection Model Traffic Light System
 30

device 1. Next will be the communication device. The communication device plays

a significant role in the communication of the two logic devices. The

communication device that is used is an outdoor long-range wireless AP Comfast

CF-EW73. The next block is the receiving logic device, which is the ESP 8266.

After receiving instructions, traffic light 2 shows a traffic signal.

Figure 6 shows the overall system operation from input to output. It also

includes the method of powering the logic devices utilized in this research. The

center of the system is the main logic device that executes instructions for all the

components, including logic device 2. This indicates that all the components

beside logic device 1 are the receivers of instructions. This is important to show to

the users of the system because it is the blueprint of the system. If troubleshooting

is performed, it will be easier.

Furthermore, the flow chart of a smart traffic light system using an object detection

model is shown in Figure 7. The flow chart will serve as a diagram to show the

logic running in logic device 1 and logic device 2. To further explain the flow chart,

the system will use an object detection model to recognize vehicles. But it will not

execute the instruction accurately if one of the components, like the camera, is not

functioning. Hence, the researchers added a feature to have a pre-testing phase

of components for detecting malfunctions. If the components are all working, it will

proceed to program execution. In the program execution, the logic device 1

(Raspberry Pi 4) will be responsible for recognizing vehicles and the execution of

logic instructions. There are two conditions that can trigger the traffic signal

change. The first one is if the road 1 (Callejon 3) logic device recognizes four or
 31

Figure 7
Flow Chart of Object Detection Model Traffic Light System
 32

more vehicles. The second condition is a combination of two conditions. These two

conditions must be met both to have a result of yes. The first condition is if the road

2 go duration is greater than 50 seconds. Second is if the road 1(Callejon 3)

recognized vehicle is not zero.

In Figure 7, it shows the flowchart of the logic condition running in the

system. This is important because some of the users require it to fully understand

the decision pattern of the system. According to Rila (2021), a flowchart is a

diagram depicting a process, a system, or a computer algorithm. This illustrates

the importance of a flowchart in presenting the system. A flowchart depicts the

process of the system for people who are not technologically oriented. For this

reason, the researchers included the flowchart shown in Figure 7.

Research Instruments

In this study, the researchers designed and developed an ODM Traffic Light

System. This traffic light system aims to solve the unorganized traffic system in the

research location. This traffic light system is also incorporated with deep learning

model specifically Yolo V5 N6 model that automatically detects vehicles within its

detection range. The detection range is under the feature of computer vision, which

is a camera. Also, the system has the capability to communicate wirelessly. The

purpose of wireless communication is to communicate between the two separate

traffic lights. These two traffic lights have a distance that is not desirable for wired

communication, so as a solution, the researchers implemented a wireless

communication system. All of the aforementioned features are just components.

To make them work together, the researchers included in the system two powerful
 33

logic devices that fit specifically to the requirements. These logic devices have the

job of performing logical processing to produce output, which is then produced as

a light by utilizing LED lights.

The researchers made a record sheet to document the results of testing the

performance of the object detection model inside the device. The record sheet

contains IoU and precision-recall curve. Based on an article written by Steen

(2020), the precision-recall curve is used for evaluating the performance of binary

classification algorithms. Intersection over Union (IoU) is used to evaluate the

performance of object detection by comparing the ground truth bounding box. The

IoU served as a guide for the researchers in evaluating the placement of the

bounding box. For this reason, the researchers included this in their documentary

analysis. Moreover, a record sheet was utilized for the purpose of recording the

speed of communication between logic devices 1 and 2. This is important because

the speed of communication in this system can affect the efficiency of delivering

output.

Range Verbal Interpretation

0.80 – 1.00 Very Much Efficient

0.60 – 0.79 Much Efficient

0.40 – 0.59 Efficient

0.20 – 0.39 Slightly Efficient

0.00 – 0.19 Not Efficient

To evaluate the object detection model's efficiency, the researchers utilized

the IoU interpretation of Rosebrock. Originally, it consisted of three interpretations

 34

of scores, but this was modified by the researchers to expand the score

interpretation. The scores that were acquired from the documentary analysis are

interpreted by the score metric evaluation below.

The researchers will utilize a questionnaire checklist to gather data from

respondents. There are three parts to the questionnaire. The first part gathers the

personal information of respondents, which consists of name, age, address, and

vehicle category. The respondents can skip this part. The second part is the

determination of the level of acceptability of the system in terms of its functional

suitability, reliability, adaptability, security and maintainability. Next is the third part,

where the researchers will gather the comments and suggestions of the

respondents about the system.

Additionally, the researchers designed a questionnaire checklist in

accordance with ISO/IEC 25010. This aims to determine the quality and

acceptability of this research. The ISO/IEC 25010 has eight characteristics, but in

this research, it will only include size: functional suitability, performance efficiency,

reliability, security, and maintainability.

Rating Scale Range Verbal Interpretation

5 4.20 - 5.00 Very Much Acceptable

4 3.40 - 4.19 Much Acceptable

3 2.60 - 3.39 Acceptable

2 1.80 - 2.59 Slightly Acceptable

1 1.00 - 1.79 Not Acceptable

To determine the result of the research, the researchers will use a 5-point
 35

Likert Scale questionnaire checklist shown which includes the scale and the verbal

interpretation above as an instrument to interpret the data gathered from the

survey.

Procedure and Method of the Study

As to construct the ODM Traffic Light System, the researchers divided into

two process: A. Conduct of the Study; and B. Conduct of the Design Development.

A. Conduct of the Study

The researchers followed a step-by-step procedure to make the research

systematic. The following steps were undertaken

The researchers came up with a possible topic and issue to submit to the

instructor for approval. During several discussions, the researchers developed the

study of ODM Traffic Light System. After the approval, the researcher started

gathering information and related studies through internet, book and library

research. By gathering all of the data, the researchers checked in the data to create

Chapters 1 and 2.

The researchers revised the chapters 1 and 2 after having consulted and

gathered opinions from the experts in order to ensure the system's complete

effectiveness. After the revision, they will submit their final draft. A colloquium oral

defense was held to evaluate the first version of this manuscript.

There was another revision after the colloquium, in which the researcher

included the expert's thinks and suggestions. Following the addition of the expert

advice and recommendations. After searching at and studying the components

that were needed, the researchers developed the design of the study and the
 36

system. The researchers also looked at the software that will be used with the

design. Furthermore, the system was tested to determine whether it was working

efficiently.

The collection of survey questionnaires was provided by the researcher in

order to gather data and feedback in order to evaluate the device's effectiveness.

After the device's creation and system's development, researchers begin work on

chapters 3 and 4.

After drafting chapters 3 and 4, the final defense was held. The panel has

recommended additional revisions to the manuscript. The critical reader analyzed

the revised manuscript. The dean gave approval for the final manuscript form to

be hardbound.

B. Conduct of the Design Development

The development of the device includes the designing, building and testing

stage.

In the design stage, the parts of the device and its information are

determined. The device consists of a Raspberry Pi, an RPI V2 camera, a Comfast

EW73, and a Node MCU (ESP8266). It also covers the outline of the design, the

features of the device, the circuit design, the algorithm of the program, and the

flowchart of the program. In addition, the researchers gathered a labeled and

trained dataset for the next stage.

Next is the building stage, where the researchers began developing the

device's software. In this stage, the researchers designed the hardware base

based on its specified design and function. This is also the stage where the
 37

researchers trained a custom object detection model. This custom training of the

object detection model was done by feeding labeled images to the base object

detection model. After the training, the researchers developed a program based

on the hardware specification and custom object detection model. Numerous tests

and adjustments were made specifically to the timing to make sure that the system

did not cause delays.

Finally, in the last stage, the researchers test the design and functionality of

the device. Test evaluations were carried out by the researchers to ensure that the

device followed the specified device features. Furthermore, in this stage, the

researchers evaluate the performance of the object detection model generated

using the documentary analysis. The problems and bugs that arose in this stage

are fixed by the researchers.

Project Design Models

The researchers utilized these metrics to evaluate the object detection

model. These metrics ensure that the researchers are not neglecting the accuracy

of the object detection.

To evaluate the object detection model in this study, the researchers utilized

three metrics. The first metric is the IoU, or intersection over union. The second

metric is recall, and the third is precision. These metrics are used in this research

in order to compute and evaluate the object detection model.

The intersection of union (IoU) is the overlap of the predicted bounding box

coordinates with the ground truth box. The higher the IoU, the higher the accuracy

of the model. In the numerator, the area of overlap is the area where the bounding
 38

box and ground truth are overlapping. On the other hand, the area of union is the

area where the bounding box and ground truth are uniting. For instance, the IoU

or ODM Efficiency score of cars. It got a 0.91974 efficiency score. To compute this,

it requires the x and y coordinates of each top left and bottom right edge of the

predicted bounding box and ground truth. The ground truth of the IoU has x and y

coordinates of 211,266,788,799 (x1,y1,x2,y2) while the predicted bounding box

has coordinates of 220,240,780,795 (x1,y1,x2,y2). To get the area of the bounding

box and the predicted box, it requires the height and width. This can be acquired

by using x2-x1 for width and y2-y1 for height. Multiplying the width and height

results in the area of each box. The area of overlap can be determined by first

finding the maximum value of x2, and y2 and the minimum value of x1 and y1.

Then next, subtract the minimum x1 from the maximum x2 and the minimum y1

from the maximum y2. Multiply the two results to produce an area of overlap. On

the other hand, area of union were computed by adding the area of the predicted

bounding box and ground truth minus the area of overlap. Dividing the area of

overlap by the area of union results to IoU.

𝐴𝑟𝑒𝑎 𝑜𝑓 𝑂𝑣𝑒𝑟𝑙𝑎𝑝
𝐼𝑜𝑈 =
𝐴𝑟𝑒𝑎 𝑜𝑓 𝑈𝑛𝑖𝑜𝑛
Furthermore, the researchers utilized the confusion matrix. According to

Bhandari (2023), a confusion matrix is a performance evaluation tool in machine

learning, representing the accuracy of a classification model. In this matrix, there

are four attributes, which are true positives, true negatives, false positives, and

false negatives.
 39

The two remaining metrics are recall and precision. Recall measures the

true positive out of all predictions. On the precision, it measures the true positives

out of all positive predictions. For instance the classification of jeepney, to get the

recall, add true positive and false negative, then use it as a divisor to divide true

positive. The jeepney has a total of 542 true positive and 107 false negative. Using

the formula, the researchers acquired 0.82 recall. The true positive and false

negative of jeepney can be acquired in the confusion matrix table in Appendix G.

True positives are classifications having matched prediction and actual

classification. On the other hand, false negatives are classifications whose actual

classification is true but the prediction is false. For example, the actual class is a

ball, but the prediction states that it is not a ball.

Moreover, the precision is computed by adding true positive and false

positive then use it as a divisor to divide true positive. For instance, the

classification of Jeep has 0.88 precision. This result can be obtained by adding

542 true positives and 68 false positives, then using it to divide 542 true positives.

𝑇𝑃
𝑅𝑒𝑐𝑎𝑙𝑙 =
𝑇𝑃 + 𝐹𝑁

𝑇𝑃
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 =
𝑇𝑃 + 𝐹𝑃

The propagation delay is being monitored by the researcher in order to

evaluate the speed and accuracy of data transmission between Raspberry Pi 4b

and Node MCU ESP8266 through Comfast EW73 access point.

 40

Propagation delay is how long it takes a signal to travel over a network from

its sender to its receiver. As computers send signals between one another, there

are delays based on the distance between the two routers.

Propagation speed is usually close to the speed of light (2.998 x 10^8 m/s)

when a signal is traveling through the air, such as a wireless signal or a signal

traveling to a satellite.

The distance between routers is the distance between the sending and

receiving router. The propagation speed is equal to the speed of light × velocity

factor. The speed of light is 2.998 x 10^8 m/s and the velocity factor ranges

between 0-1 depending on the medium through which the signal is propagating.

Moreover, in this research, the distance of between routers is 72 meters. Dividing

72 meters by the speed of light resulted in a computed propagation delay of

0.00000024016 seconds. This computed propagation delay serves as the base

comparison for the propagation delay acquired in testing.

Statistical Treatment

The researchers utilized the weighted mean, average, standard deviation,

and Intersection over Union formula. These formulas were used to determine the

interpretation of the gathered data. Additionally, it also helps the researchers

present the data in a compact and comprehensible form.

 41

To test the performance of the developed Object Detection Model Traffic

Light Systems, the researchers utilized Intersection over Union and mean formula.

The Intersection over Union was specifically used to compute the performance of

each vehicle category. Additionally, the researchers incorporated it with mean and

standard deviation to compute the overall performance of the object detection

model and propagation delay.

To determine the level of acceptability of the developed Object Detection

Model Traffic Light System in terms of functional suitability, compatibility, reliability,

usability, security, and maintainability, weighted mean was used.

See Appendix G for the statistical formula used.

 42

Chapter 3

PRESENTATION, ANALYSIS, AND DESIGN RESULTS

This chapter presents analysis and interpretation of data on the

development of the Object Detection Model Traffic Light System. Likewise, this

chapter discusses the level of the acceptability of the system and the performance

of the Object Detection Model Traffic Light System.

Design and Development of an Object Detection Model Traffic Light System

The research design was based on developmental processes in designing,

developing, and evaluating instructional programs, processes, and products that

must meet the criteria of internal consistency and effectiveness. Additionally,

researchers utilized the descriptive methods of descriptive research to answer the

questions for this research. The researchers used the innovation life cycle to

discuss the development of the system.

The Object Detection Model Traffic Light System was designed and

developed in accordance with its purpose and function. The main purpose of the

study is to provide a traffic light system to manage the traffic in Philsin Bridge,

Barangay Dalig, Teresa, Rizal.

In figure 8, it shows the RPi camera module that was used in this study to

assist the RPi 4 microcontroller in object detection, which can be seen in three

dimensions in the picture below.

In this study the researchers used an RPi camera module to serve as eyes

for the AI so that it can detect objects and is also compatible for the RPi

microcontroller.
 43

This research investigates the utilization of the Raspberry Pi Camera

Module as a pivotal tool in refining Object Detection Model Traffic Light System.

The compact and versatile Raspberry Pi Camera Module serves as the sensory

component, capturing real-time visual data crucial for detecting vehicles at

intersections. Integrating advanced object detection algorithms with the Raspberry

Pi Camera Module facilitates the accurate identification and classification of

objects, enabling efficient traffic flow management and enhancing overall safety

on roadways.

Figure 8
RPi camera V2 (Computer Vision)

The researchers applied the RPi camera module, by integrating this

compact camera module, and the study aims to enhance the system's ability to

detect vehicles at intersections, thereby improving traffic flow and overall safety on

the roads.

In the context of a research title focusing on object detection within traffic

light systems, the Raspberry Pi Camera Module plays a crucial role in integrating
 44

computer vision capabilities. As outlined by Ashtari (2023), computer vision

leverages artificial intelligence (AI) to enable computers to extract valuable

information from visual inputs, including photos and videos. This research

capitalizes on computer vision, facilitated by the Raspberry Pi Camera Module, to

enhance the system's ability to perceive and identify objects such as vehicles and

pedestrians at traffic intersections. By harnessing the power of computer vision,

the system can take automated actions, ultimately contributing to improved traffic

management and safety.

Figure 9 below shows the result of train batch 1 of the object detection

model. In this figure, it shows vehicles that have boxes on each of them.

Figure 9
Yolo V5 N6 (Object Detection Model)

These boxes also have different colors. These boxes are called bounding

boxes. The bounding box is the box placed on a picture or frame of video to show

the detection of the object, like a vehicle. Also. in this figure, these boxes have
 45

different colors. This differentiation in colors is an illustration of classification. Each

color of the box has a different vehicle equivalent.

Additionally, figure 9 shows the results of the training of the object detection

model. The object detection model utilized in this research was Yolo V5 N6. This

object detection model was the most optimized version of Yolo, but to make it more

efficient, real-time, and accurate, the researchers trained the model to specifically

detect vehicles in the Philippines. The results of the training are shown in the figure

above.

According to Kundu (2023), object detection deals with localizing a region

of interest within an image and classifying this region like a typical image classifier.

One image can include several regions of interest pointing to different objects. This

makes object detection a more advanced problem of image classification. Yolo is

a deep learning model that classifies objects in a video frame by frame. The

aforementioned article shows the process of detection in an objection model. This

explains how the object detection model in this research can detect multiple

vehicles in one frame. This explains how the object detection model in this

research can detect multiple vehicles in one frame. This feature of object detection

together with the real-time detection of Yolo is the reason why the researchers

chose object detection over analog sensors.

The figure 10 below shows a 3D model of a Raspberry Pi 4 microcontroller

that was used in this study for AI detection.

In this study, the Raspberry Pi 4, a multipurpose single-board computer, is used

as a foundation for improving object identification models in traffic light systems.

 46

By utilizing the Raspberry Pi 4's small size and powerful capabilities, this research

seeks to improve traffic signal systems' accuracy and efficiency by implementing

smart object detection algorithms.

Figure 10
Raspberry Pi Microcontroller (Logic Device 1)

In this study, the Raspberry Pi 4, a multipurpose single-board computer, is

used as a foundation for improving object identification models in traffic light

systems. By utilizing the Raspberry Pi 4's small size and powerful capabilities, this

research seeks to improve traffic signal systems' accuracy and efficiency by

implementing smart object detection algorithms.

The researchers were able to efficiently recognize items with the Raspberry

Pi 4 microcontroller in our research by utilizing a data set to teach the AI what

objects to look for on the RPi camera module. While employing this microcontroller,

the researchers also ran into some issues, but they quickly fixed them. Among

these challenges were issues with the access point's internet connection and the

other devices' wire connection.

The Raspberry Pi 4, a compact yet powerful single-board computer, serves

 47

as a cornerstone for innovative projects, particularly in the realm of artificial

intelligence (AI). Frankenfield (2023) defines AI as the simulation of human

intelligence through software-coded heuristics, a concept prevalent in modern

computing. Leveraging the Raspberry Pi 4's robust processing capabilities,

researchers utilize AI to develop object detection models for traffic light systems.

This integration allows the Raspberry Pi 4 to recognize and categorize objects,

enabling sophisticated traffic management solutions. In this study, the Raspberry

Pi 4 embodies the fusion of AI and embedded firmware, showcasing its versatility

in real-world applications.

The figure 9 below shows a 3D model of a COMFAST EW73 WiFi Module

that was used in this study as an access point for the communication of the

microcontrollers.

Figure 11
Comfast EW73 WiFi Access Point (Wireless Communication)
 48

This research investigates the potential of incorporating the COMFAST

EW73 WiFi Module alongside microcontrollers Node MCU and Raspberry Pi 4 to

enhance object detection within traffic light systems using the internet of things.

The researchers devised a method to enable communication between the

two microcontrollers despite their distance from one another. The two

microcontrollers can now link to each other with a point-to-point connection with

the use of the COMFAST EW73 WiFi Module, enabling them to signal each traffic

light synchronously using the internet of things.

In traffic light system research, the COMFAST EW73 Access Point is

essential for enabling wireless connectivity, facilitating the integration of smart

object identification models. Yasay (2021) highlights the role of the Internet of

Things (IoT) in connecting virtual and physical entities, utilizing technologies like

embedded systems, real-time analytics, and machine learning. To streamline

connections and enhance efficiency, wireless communication is imperative. The

COMFAST EW73 WiFi Module is chosen for its extensive communication

capabilities, crucial for traffic control systems. This study aims to optimize object

identification in traffic light systems using Wi-Fi, specifically leveraging the

COMFAST EW73, to improve roadway efficiency and safety while overcoming

wired connection limitations.

The figure 11 illustrates the 3D model of the Node MCU ESP8266. The

Node MCU ESP8266 was used as a secondary logic device that receives

instructions from the primary logic device, which is the Raspberry Pi 4B. This

component was specifically chosen by the researchers because of its features.

 49

Figure 12
Node MCU ESP8266 (Logic Device 2)

The figure above shows the Node MCU ESP8266. This device was utilized

by the researchers in order to make the traffic light 2 capable of connecting to a

wireless connection. However, the researchers found difficulty connecting to the

wireless connection. The researchers found out that wireless connections can be

interrupted by metallic objects. For this reason, the researchers revised the design

by adding an antenna connected to the built-in antenna of Node MCU ESP 8266.

Hence, according to Sonawane et al., (2017), the very slow speed of the network

leads to limited connectivity and delays in processing hours, so people tend to use

wireless technology either at home, at a coffee shop, or at the airport. This article

proves that delays can reduce work output. The same problem was faced in this

research, where the output of the device was reduced because of the delay caused

by slow connectivity. Also, same with the idea of the article, the device in this
 50

research finds a way to connect more efficiently to the wireless network, and this

is by adding an antenna to the device.

The figure shows a 3D model of a Traffic Light that was used as the main

design or model in this study.

Figure 13

Light Emitting Diode (Traffic Signal Light)

This study explores how state-of-the-art object detection models can be

applied to modify conventional traffic lights. The goal of the project is to maximize

traffic flow and safety by incorporating cutting-edge technologies like computer

vision and artificial intelligence into traffic signal systems. The goal of the project

is to improve traffic light responsiveness and efficiency in identifying cars by using

object detection algorithms. This innovative approach aims to transform traffic

management, which will eventually result in more efficient traffic flow and

increased road safety.

The researchers used a traffic light as a model for this study. Focusing on

the problem that the researchers saw at Pahati Bridge, which was traffic
 51

congestion. The researchers thought of a way to solve the problem, and that was

to use a traffic light at Pahati Bridge to minimize traffic congestion in that area.

The traffic light is a crucial element of the traffic management system when

it comes to tackling the pressing problem of traffic congestion. According to Fattah

et al. (2021), congestion has a negative influence on both the economy and road

users, as well as having a substantial economic and social impact. Researchers

are looking into creative ways to solve this issue, such as incorporating object

detection models into traffic signal systems. Through the utilization of cutting-edge

technology, this study complies with Republic Act No. 4136, Section 35, which

requires careful driving and safety precautions on public roads. In keeping with the

transportation regulations, object detection models added to traffic lights not only

improve traffic efficiency but also increase overall road safety.

The researchers used a traffic light as a model for this study. Focusing on

the problem that the researchers saw at Pahati Bridge, which was traffic

congestion. The researchers thought of a way to solve the problem, and that was

to use a traffic light at Pahati Bridge to minimize traffic congestion in that area.

The traffic light is a crucial element of the traffic management system when

it comes to tackling the pressing problem of traffic congestion. According to Fattah

et al. (2021), congestion has a negative influence on both the economy and road

users, as well as having a substantial economic and social impact. Researchers

are looking into creative ways to solve this issue, such as incorporating object

detection models into traffic signal systems. Through the utilization of cutting-edge

technology, this study complies with Republic Act No. 4136, Section 35, which
 52

requires careful driving and safety precautions on public roads. In keeping with the

transportation regulations, object detection models added to traffic lights not only

improve traffic efficiency but also increase overall road safety.

Test the Performance Efficiency of the Developed Object Detection Model Traffic
Light System

To test the performance of the Object Detection Model, the researchers

utilized the Intersection over Union, Recall, and Precision formula. The

aforementioned formulas are the tools to determine the efficiency and accuracy of

the object detection model utilized.

Table 1 presents the Computed Mean of the developed system Object

Detection Model Traffic Light System.

The gathered and computed mean below is interpreted using the IoU

evaluation metrics.

Table 1

Computed Mean of Object Detection Model Efficiency of the developed

Object Detection Model Traffic Light System

Vehicle ODM Eff. Verbal Interpretation

Tricycle 0.8145 Very Much Efficient

Jeepney 0.8971 Very Much Efficient
Truck 0.9078 Very Much Efficient
Cars 0.9200 Very Much Efficient
Van 0.8400 Very Much Efficient
Motorcycle 0.7500 Much Efficient
Bus 0.9011 Very Much Efficient
Mean (x̄) 0.8615 Very Much Efficient
 53

Based on the table above, the truck, bus, van, tricycle, cars, and jeepney

acquired excellent scores. On the other hand, motorcycles have the lowest score.

These data imply that the object detection model's accuracy is interpreted as

excellent.

The researchers find out that the most efficient classification of vehicle is

“Cars” with a score of 0.92 and an interpretation of “Very Much Efficient”. On the

other hand, the lowest score in “ODM Efficiency” is the “Motorcycle” with a score

of 0.75 and an interpretation of “Much Efficient”. This table illustrates that the

efficiency of detection is proportional to the rarity, and size of the vehicle. The rarity

of the vehicle affects its detection efficiency, because of the datasets. The lack of

a labeled dataset sample of the specific object can affect its detection efficiency.

Additionally, the pixel and camera conditions affect the detection of the object

detection model. Bigger objects are easily detected because of their pixel quality

and clearer boundaries. As stated by Kundu (2023), object detection deals with

localizing a region of interest within an image and classifying this region like a

typical image classifier. One image can include several regions of interest pointing

to different objects. This makes object detection a more advanced problem of

image classification. This article shows the compatibility of the YOLO with road

vehicle detection. The researchers decided that its accuracy, speed, and multi-

object detection capability are applicable for this study. This decision resulted

great, as shown in the table above.

Table 2 presents the Propagation delay of the developed system Object

Detection Model Traffic Light System.

 54

Table 2

Computed Mean of Propagation Delay of the developed

Object Detection Model Traffic Light System

Time Propagation Delay

Early morning (5 - 8 am) 0.04230 seconds

Late morning (11am - 0.00024 seconds

12pm)
Early afternoon (1 - 3pm) 0.05600 seconds

Late afternoon (4 - 5pm) 0.95000 seconds

Early evening (5 - 7 pm) 0.89000 seconds

Night (9 pm - 4 am) 0.00667 seconds

Mean (x̄) 0.32420 seconds

Propagation delay is influenced by the speed of signal propagation in the

medium as well as the distance between the source and the destination.

Propagation delay is critical in the design and analysis of communication systems

because it affects the system's overall performance and dependability, particularly

in time-sensitive applications such as telecommunications and networking.

This table shows the propagation delay lengths measured at different times

of the day. During the early morning, late morning, and night, the delay is minimal,

the lowest propagation delay is observed during the "Late morning (11am - 12pm)"

time slot, with a delay of 0.00024 seconds. This minimal delay could be attributed

to several factors. Late morning typically experiences lighter traffic compared to

peak hours, resulting in simpler traffic scenarios for the object detection system to

handle.
 55

On the other hand, the delay increases significantly in the late afternoon

and early evening, reaching up to 0.95 seconds and 0.89 seconds, the highest

propagation delay in the table occurs during the "Late afternoon (4 - 5pm)" time

slot, with a delay of 0.95 seconds. This could be justified by considering the

increased traffic and activity during this time period. These fluctuations are most

likely the result of network congestion, usage patterns, and infrastructure capacity,

all of which have an influence on communication efficiency during the day.

Based on an article written by Sonawane et al., (2017), the delays in the

Wi-Fi network might be caused either by server failure or by a loss of range. They

also mentioned that this lost connection is caused by Wi-Fi dead zones. This article

made the researchers design the position and location of the devices strategically,

making wireless communication seamless. This method reduced the chance of

larger delays that might affect the efficacy of the traffic light in the research

location.

Level of Acceptability of the Developed Object Detection Model Traffic Light

System

The researchers were able to gather the acceptability of the project

according to the functionality suitability, performance efficiency, portability,

reliability, security, and usability. This will show the result how acceptable the

device is.

The following tables below shows the result of the computed level of

acceptability of the goals of the study.

Table 3 presents the computed weighted mean on the level of acceptability

computed Weighted Mean on the Level of Acceptability of the Object Detection

 56

Model Traffic Light System in terms of Functional Suitability.

Table 3

Computed Weighted Mean on the Level of Acceptability of the developed Object

Detection Model Traffic Light System in Terms of Functional Suitability

1. Functional Suitability Respondents

The developed device… Wx̄ Verbal Interpretation
1.1 effectively adjusts to variations in 4.70 Very Much Acceptable
lighting conditions, weather, and
other environmental factors.
1.2 consistently achieves accurate 4.50 Very Much Acceptable
detection of vehicles.
1.3 has contributed to a reduction in road 4.54 Very Much Acceptable
rage incidents in the area.
1.4 prevents traffic congestion in the 4.60 Very Much Acceptable
area.
1.5 The signal timing implemented in the 4.52 Very Much Acceptable
system is efficient.
4.58 Very Much Acceptable
Average

It can be gleaned in the table that with respect to functional suitability, the

respondents agreed that the developed system is Very Much Acceptable with an

overall weighted mean of 4.58 from the result. Item 1.1 “effectively adjusts to

variations in lighting conditions, weather, and other environmental factors.” with a

weighted mean of 4.70 and verbally interpreted as Very Much Acceptable.

It implies that the system’s functional suitability fulfil accordingly and satisfy

the desired of the respondents. It means that the system that was developed

performs in accordance with its function.

According to Widartha et al., (2023), increasing demand for energy-efficient

and sustainable indoor environments has led to the development of various smart

lighting and energy consumption management systems. In their study, they

 57

developed a smart lighting system that can efficiently reduce electricity usage. This

was done by making a system integrated with a sensor to sense the amount of

light entering the room. Additionally, they added daylight data to accurately adjust

the lighting. The system they made can make energy consumption more efficient

by adjusting the light. Like the system in this research, its adaptability can produce

more efficient results. The reason behind this is that the system can be used

anytime, anywhere, and in any condition.

Table 4 Presents the computed weighted mean on the level of acceptability

computed Weighted Mean on the Level of Acceptability of the Object Detection

Model Traffic Light in terms of Compatibility.

Table 4

Computed Weighted Mean on the Level of Acceptability of the developed Object

Detection Model Traffic Light System in Terms of Compatibility

2. Compatibility Respondents
The developed system… Wx̄ Verbal Interpretation
2.1 is compatible with any USB type 4.51 Very Much Acceptable
camera.
4.61 Very Much Acceptable
2.2 is applicable to other location.

2.3 is transferable to a new administrator. 4.51 Very Much Acceptable

2.4 The logic device of the developed 4.63 Very Much Acceptable
system can be improved to achieve better
performance in detection.
2.5 The dataset used in the system is 4.49 Very Much Acceptable
upgradable to prevent inaccurate
detection.
Very Much Acceptable
Average 4.55

It can be seen from the table that with respect to compatibility, the

respondents agreed that the developed system is Very Much Acceptable with an
 58

overall average weighted mean of 4.55. Item 2.4, “The logic device of the

developed system can be improved to achieve better performance in detection.” is

rank first with a weighted mean of 4.63, whereas item 2.2 “is applicable to other

location.” With a weighted mean of 4.61. While 2.1 “is compatible with any USB

type camera.” and 2.3 “is applicable to other location.” Were ranked last with both

having a weighted mean of 4.51 also interpreted as Very Much Acceptable.

It can be realized from the findings that the system’s compatibility fulfils the

desires of the respondents. This only means that the system developed satisfied

the respondents.

According to Feng et al., (2023), they found out that the Jetson Xavier

outperformed the Raspberry Pi 4B and Jetson Nano in running Yolo V3. Also, they

noticed that the highest performing device also has the highest power

consumption. This proves the result of the table above. The Raspberry Pi 4B can

be replaced with a more powerful logic device, but together with the increase in

processing power, the power consumption and price of the system will also

increase. For this reason, the researchers utilized the Raspberry Pi 4B as the main

logic device.

Table 5 presents the computed weighted mean on the level of acceptability

computed Weighted Mean on the Level of Acceptability of the Object Detection

Model Traffic Light System in terms of Reliability.

It can be gleaned from the table that in terms of reliability, the respondents

agreed that the developed system is Very Much Acceptable with an overall

average weighted mean of 4.53.

 59

Table 5

Computed Weighted Mean on the Level of Acceptability of the developed Object

Detection Model Traffic Light System in Terms of Reliability

3. Reliability Respondents
The developed system… Wx̄ Verbal Interpretation
3.1 consistently provides accurate results 4.54 Very Much Acceptable
in detecting vehicles.
3.2 performance is reliable over extended 4.49 Very Much Acceptable
periods of continuous operation.
3.3 reliably detects vehicles under 4.56 Very Much Acceptable
challenging conditions, such as low
visibility or adverse weather.
3.4 can return to normal operation when 4.51 Very Much Acceptable
there is a power interruption.
3.5 Users can depend on the developed 4.57 Very Much Acceptable
system to consistently deliver accurate
information about traffic light status.
4.53 Very Much Acceptable
Average

Item 3.5, “Users can depend on the developed system to consistently

deliver accurate information about traffic light status.” Is rank first with a weighted

mean of 4.57. Whereas item 3.3 “reliably detects vehicles under challenging

conditions, such as low visibility or adverse weather.” with a weighted mean of

4.56. Item 3.1 “consistently provides accurate results in detecting vehicles.” with a

weighted mean of 4.54. Item 3.4 “can return to normal operation when there is a

power interruption.” With a weighted mean of 4.51. Lastly item 3.2 “performance is

reliable over extended periods of continuous operation.” Has a weighted mean of

4.49, all of the items are “Very Much Acceptable”.

It can be noted from the result that the system’s reliability fulfils the desires

of the respondents. It means that the system can be used in other places.
 60

According to an article written by Umali (2019), there are students who have

developed software that would help authorities quickly and more efficiently identify

traffic violators. CATCH-ALL, or the Contactless Apprehension of Traffic Violators

on a 24-Hour Basis and All-Vehicle Detection System, is capable of detecting and

tracking vehicles. As stated in the table above, the system in this research can

also efficiently perform its specified functions, so the users can rely on it even for

critical tasks.

Table 6 presents the computed weighted mean on the level of acceptability

computed Weighted Mean on the Level of Acceptability of the Object Detection

Model Traffic Light System in terms of Usability.

Table 6

Computed Weighted Mean on the Level of Acceptability of the developed Object

Detection Model Traffic Light System in Terms of Usability

4. Usability Respondents
The developed system… Wx̄ Verbal Interpretation
4.1 users can easily understand and 4.59 Very Much Acceptable
interpret the information provided by the
developed system.
4.2 training and utilizing the dataset require 4.54 Very Much Acceptable
minimal effort and technical expertise.
4.3 operating the developed system 4.57 Very Much Acceptable
requires minimal effort or technical
expertise.
4.4 the website controlling the power of the 4.63 Very Much Acceptable
traffic light is easy to operate and
understand
4.5 does its job of guiding the drivers 4.60 Very Much Acceptable
based on the requirements of the location.
4.59 Very Much Acceptable
Average

It could be gleaned on the table that the observation of the respondents

when it comes to usability item 4.4 “the website controlling the power of the traffic
 61

light is easy to operate and understand.” got the highest mean of 4.63, next is item

4.5 “does its job of guiding the drivers based on the requirements of the location.“

with weighted mean of 4.60, item 4.1 “users can easily understand and interpret

the information provided by the developed system.” with 4.59 weighted mean.

Second to the last is the item 4.3 “operating the developed system requires

minimal effort or technical expertise.” with weighted mean of 4.57. Lastly the item

4.2 “training and utilizing the dataset require minimal effort and technical

expertise.” which got the lowest mean of 4.54, and all the items are” Very Much

Acceptable”.

The Respondents agreed that the developed system is Very Much

Acceptable with an overall weighted mean of 4.59 from the result, it implies that

the system’s usability fulfil accordingly and satisfy the desired of the respondents.

According to Corpus et al., (2023), there are still some traffic symbols that

have a noticeable rate of inaccurate identification by drivers. This was caused by

the design. They also emphasize that there is room for improvement in visual

design. In their recommendation, they said that authorities could provide

educational training to increase traffic sign knowledge. However, this educational

seminar is not needed if the design was not complicated in the first place. In

contrast to the traffic signs, the website interface that controls power shows a great

result in the rate of user-friendliness of the interface.

Table 7 Presents the computed weighted mean on the level of acceptability

computed Weighted Mean on the Level of Acceptability of the Object Detection

Model Traffic Light System in terms of Security.

 62

Table 7

Computed Weighted Mean on the Level of Acceptability of the developed Object

Detection Model Traffic Light System in terms of Security

5. Security Respondents
The developed system… Wx̄ Verbal Interpretation
5.1 includes strong security measures to 4.40 Very Much Acceptable
prevent unauthorized access.
5.2 sufficient safeguards are implemented 4.39 Very Much Acceptable
to ensure the privacy of the information
handled by the system.
5.3 the casing of the device is designed to 4.34 Very Much Acceptable
protect the components inside from
attackers.
5.4 the accounts and passwords used in 4.49 Very Much Acceptable
this system are stored in a safe place.
5.5 the communication system used 4.40 Very Much Acceptable
between logic boards are secured.
Average 4.40 Very Much Acceptable

Table 7 shows the observation of the respondents when it comes to security

item 5.4 “the accounts and passwords used in this system are stored in a safe

place” which got the highest mean of 4.49, next is item 5.1 “includes strong security

measures to prevent unauthorized access.” and 5.2 “the communication system

used between logic boards are secured. “got the highest same mean of 4.40.

Lastly is item 5.2 “the casing of the device is designed to protect the components

inside from attackers.” got the lowest mean with 4.34, and all items interpreted as

“Very much acceptable”.

The Respondents agreed that the developed system is Very Much

Acceptable with an overall average weighted mean of 4.40 from the result it implies

that the system’s Security fulfil the desires of the users.

 63

In an article written by Blue et al., (2017), there are numerous ways to

protect database storage that stores passwords. They mention the hashing, where

the passwords are hashed to protect them even if attackers have access to the

database. Another security measure is to provide a password policy that enforces

length, entropy, and lifespan. This may provide safe guards against password

guessing or brute-force attacks. In this research, the system was protected first in

the network. This was done by blocking unknown IP addresses and Mac

addresses. Additionally, the password's database storage is protected inside its

memory. This preventive action ensures the security of the storage of passwords.

This was reflected in the results shown in the table above.

As shown at table 8, Presents the computed weighted mean on the level of

acceptability computed Weighted Mean on the Level of Acceptability of the Object

Detection Model Traffic Light System in terms of Maintainability.

Table 8

Computed Weighted Mean on the Level of Acceptability of the developed Object

Detection Model Traffic Light System in terms of Maintainability

6. Maintainability Respondents
The developed system… Wx̄ Verbal Interpretation
6.1 the parts of the developed system are 4.40 Very Much Acceptable
affordable.
6.2 requires less effort to troubleshoot 4.31 Very Much Acceptable
because it tests itself before operation.
6.3 troubleshooting hardware components 4.39 Very Much Acceptable
of the device is convenient because the
parts and wires are properly organized.
6.4 the parts of the developed system are 4.31 Very Much Acceptable
accessible.
6.5 the developed system is convenient to 4.31 Very Much Acceptable
move.
Average 4.56 Very Much Acceptable
 64

As shown at table 8, it shows that based on the observation of the

respondents when it comes to maintainability item 6.1” the parts of the developed

system are affordable ”which got the highest mean of 4.40, next is item 6.3

“troubleshooting hardware components of the device is convenient because the

parts and wires are properly organized.” with 4.39 weighted mean, Lastly, item 6.2

“requires less effort to troubleshoot because it tests itself before operation.”, item

6.4 “the parts of the developed system are accessible.” And item 6.5 “the

developed system is convenient to move.” which got the lowest mean of 4.31, and

all items interpreted as “Very much acceptable”.

The respondents agreed that the developed system is Very Much

Acceptable with an overall average weighted mean of 4.56 from the result it implies

that the system’s Usability fulfil the desires of the respondents.

According to an article written by Valerio (2023), Arm Holdings plc has

announced that they have made a strategic investment in Raspberry Pi. The

Raspberry Pi was originally intended for students, but it has become a tool to

develop Internet of Things (IoT) prototypes and items. After securing their

agreement with ARM, the latter company gained a minority stake in the Raspberry

Pi. It is said in the article that IoT solutions will be more accessible to students and

anyone by lowering the barriers to innovation. For this reason, the researchers

utilize materials and devices that are affordable and accessible. This includes

Raspberry pi that is known for affordability, flexibility, and accessibility.

Table 9 shows the level of acceptability of the developed Object Detection

Model Traffic Light System. It is very much acceptable in term of Functional

 65

Suitability, Performance Efficiency, Portability, Reliability, Security and Usability.

Table 9

Composite Table of the Level of Acceptability of the developed

Object Detection Model Traffic Light System

Aspects Average Verbal Interpretation

1. Functional Suitability 4.58 Very Much Acceptable

2. Compatibility 4.55 Very Much Acceptable

3. Reliability 4.53 Very Much Acceptable

4. Usability 4.59 Very Much Acceptable

5.Security 4.40 Very Much Acceptable

6.Maintainability 4.56 Very Much Acceptable

Grand Mean 4.54 Very Much Acceptable

As observed on table 9, Usability is first rank with an average weighted

mean of 4.59. Functional suitability is second in rank with an average weighted

mean of 4.58. The Maintainability at third rank with an average weighted mean of

4.56. While the Compatibility had an average weighted mean 4.55. And also

Reliability had an average weighted mean 4.53. Lastly, Security had an average

weighted mean of 4.40 as the lowest. Overall, the result displays an average

weighted mean of 4.54 for all variables.

It simply implies that the develop Object Detection Model Traffic Light

System is very much acceptable and could state the facts in terms of Functional

Suitability, Compatibility, Reliability, Usability, Security, and Maintainability.

Based on the study written by Ecaldre et al., (2022), using pre-trained deep

learning and a color-connected component algorithm, the prototype model was

 66

successfully simulated using an image processing tool. The prototype was able to

manage traffic at a particular flow, giving priority to the green light where there

were more cards in one lane than the other. In their study, they concluded that

using pre-trained deep learning could lessen congestion and waste time at traffic

lights. In other words, their research satisfied its usability, where the system in their

research achieved goals effectively, efficiently, and satisfiably. Together with this

research, the table above shows that the usability aspect received the highest

scores from the survey.

 67

Chapter 4

SUMMARY OF DESIGN RESULTS, CONCLUSIONS, AND

RECOMMENDATIONS

This study developed an “Object Detection Model Traffic Light System”. The

respondents were Community, Local Government Unit and Experts who evaluated

the system through Questionnaire checklist after the researcher’s presentation of

it.

Summary of Design Results

The summary of the design results stated as follows were established based

on the evaluation made by the respondents.

1. With a weighted mean of 4.59, the usability result is the highest and

indicates that the system fulfils users' needs and expectations as intended. With a

weighted mean of 4.40, the security result is the lowest, indicating that users'

needs are met by the system's security. With an overall average weighted mean

of 4.56, the experts and students agreed that the established system is very much

acceptable. This suggests that the usability of the system satisfies the needs of

the users.

2. The researchers used the Object Detection Efficiency, Recall, and

Precision calculation to evaluate the Object Detection Model's performance. The

preceding formulas serve as instruments for assessing the effectiveness and

precision of the applied object detection model.

3. The researchers were able to determine whether the project was

acceptable. Based on the level of the functionality, performance, portability,

security, and usability. This will display the device's level of acceptability.
 68

Conclusions

Based on the summary of the design result the researchers therefore

conclude the following:

1. The integration of the Video Capturing Device (RPi Camera V2), Object

Detection Model (Yolo V5 Nano 6), Logic Devices (Raspberry Pi 4 and Node MCU

ESP8266), Wireless Communication (Comfast EW73), and Traffic Signal Light

(LED) forms a comprehensive system for real-time traffic monitoring and

management. This system utilizes advanced technologies for efficient object

detection, data processing, and wireless communication, enabling effective control

of traffic signals based on detected conditions. By using these components, the

system enhances traffic safety, efficiency, and overall management, contributing

to smarter and more adaptive urban environments.

2. The performance efficiency of the developed Object Detection Model

within the Traffic Light System was evaluated through two key aspects which is the

Object Detection Model itself and the Propagation Delay associated with wireless

communication. Overall, through testing of both the Object Detection Model and

the Propagation Delay validated the Traffic Light System's performance efficiency.

The system demonstrated effective vehicle detection and timely control signal

adjustments, contributing to efficient traffic management and enhanced road

safety. Ongoing monitoring and optimization efforts are recommended to maintain

and improve system performance over time.

3. The assessment of the developed Object Detection Model Traffic Light

System across various dimensions of the level of acceptability yielded positive

 69

results. The system proved highly suitable for its intended purpose, effectively

detecting vehicles within the designated camera range and adjusting traffic signals

accordingly. Overall, the Object Detection Model Traffic Light System

demonstrated a high level of acceptability across various dimensions, including

functional suitability, compatibility, reliability, usability, security, and maintainability.

Its successful evaluation underscores its effectiveness in enhancing traffic

management efficiency and safety on roadways.

Recommendations

Based on the analysis, evaluation of the design results and conclusion the

researchers recommended the following:

1. Implement regular performance monitoring, optimize the Object

Detection Model, ensure redundancy in critical components, enhance security

measures, and design for scalability to maximize the effectiveness of the

integrated system for real-time traffic monitoring and management in Pahati

Bridge.

2. Continuously monitor and optimize both the Object Detection Model and

wireless communication to ensure consistent and efficient vehicle detection, signal

adjustments, and traffic management within the Traffic Light System, fostering

sustained improvement in performance and road safety.

3. Prioritize ongoing evaluation and refinement of the Object Detection

Model Traffic Light System across all dimensions of acceptability, including

functional suitability, compatibility, reliability, usability, security, and maintainability,

 70

to ensure its continued effectiveness in optimizing traffic management efficiency

and enhancing road safety.

4. Future researchers should conduct further research on the acceptability

of the study to be able to produce another effective output from the entire system.

Future researchers should provide a detailed traffic information with use of a timer.

The future researchers should conduct and provide research on how to make the

dataset more accurate for better object detection.

 71

REFERENCES

Ashtari, Hossein, (2022). Review on “What Is Computer Vision? Meaning,

Examples, and Applications in 2022”, Retrieved from
https://www.spiceworks.com/tech/artificial-intelligence/articles/what-iscom
putervision/?fbclid=IwAR3jJwMikoOH7r4LSMr3JtGDe31hsHEUo3sq0pR5
pWarXzmXXc2RbrqxvDs, Retrieved on May 13, 2022.

Bhandari, Aniruddha, (2023). Review on “Understanding & Interpreting Confusion

Matrix in Machine Learning”, Retrieved from
https://www.analyticsvidhya.com/blog/2020/04/confusion-matrix-machine-
learning/?fbclid=IwAR0GkVyWkDW_Ldzvbv36IEUi0tBGZv6PA8DYwdzo5
y4s0XtrYw7YZIPlD98#:~:text=A%20confusion%20matrix%20is%20a%20
performance%20evaluation%20tool%20in%20machine,false%20positives
%2C%20and%20false%20negatives, Retrieved on November 02, 2023.

Blue, Juanita, Furey, Eoghan, Curran, Kevin, (2017). Review on “An Evaluation of
Secure Storage of Authentication Data”, Retrieved from
338495788_An_Evaluation_of_Secure_Storage_of_Authentication_Data,
Retrieved on June 2017.

Chablani, Manish, (2017). Review on “YOLO — You only look once, real time
object detection explained”, Retrieved from
https://towardsdatascience.com/yolo-you-only-look-once-real-time-object-
detection-explained-492dc9230006, Retrieved on August 21, 2017.

Corpus, Robert Michael, Suguitan, Daniel, Tamis, Tamis, Mark Albert, (2023).
Review on “The Impact of Autonomous Vehicles on Traffic Flow and
Congestion in the Philippines: A Comparative Study of Autonomous and
Human-Driven Vehicles”, Retrieved from
https://www.researchgate.net/publication/373714147_The_Impact_of_Aut
onomous_Vehicles_on_Traffic_Flow_and_Congestion_in_the_Philippines
_A_Comparative_Study_of_Autonomous_and_Human-Driven_Vehicles,
Retrieved on September 2023.

Delgado, R., Catahimican, A., Deciar, J., Gumisad, P. Automatic Car Parking
System with Plate Recognition Using Online Optical Character Recognition.
University of Rizal System, 2023.

Ecaldre, M., Besonaya, J., Bungay, G. Smart Traffic Light Control System Using
an Image Processing Tool. University of Rizal System, 2022.

Espinoza, Alfonso, Navarro, Bonilla, Oscar Roberto, Lopez, Guerrero, Enrique

Efren, Garcia, Cuautle, Esteban, Tlelo, Mancilla, Didier, Lopez, (2022).
Review from “Traffic Flow Prediction for Smart Traffic Lights Using Machine
Learning Algorithms”, Retrieved from https://www.sciencegate.app/docume
 72

nt/10.3390/technologies10010005?fbclid=IwAR3F_lmzDsdoDPLtySRlcCV
e_E5lCmtdhe-WGDozVTGAEMcwduEyLTeQ_7k, Retrieved on January
10, 2022.

Fattah, Md. Abdul, Morshed, Syed Riad, Kafy, Abdulla-Al, (2022). Review on
"Insights into the socio-economic impacts of traffic congestion in the port
and industrial areas of Chittagong city", Bangladesh",
Retrieved from https://www.sciencedirect.com/science/article/pii/S2666691
X22000203?fbclid=IwAR1X6J3gyFHv-mT5j34lPwWnAPk4ViEWXYJOgP-
w4E4ZfFip9WhLmilraaA, Retrieved on June 30, 2022.

Feng, Haogang, Mu, Gaoze, Zhong, Shida, Zhang, Peichang, (2022). Review on
“Benchmark Analysis of YOLO Performance on Edge Intelligence Devices”,
Retrieved from https://www.researchgate.net/publication/359689579Bench
mark_Analysis_of_YOLO_Performance_on_Edge_Intelligence_Devices,
Retrieved on April 2022.

Frankenfield, Jake, (2023). Review on “Artificial Intelligence (AI): What It Is and

How It Is Used”, Retrieved from https://www.investopedia.com/terms/
a/artificial-intelligence-ai.asp#:~:text=Artificial%20intelligence%20(AI)%20
refers%20to,industries%20from%20finance%20to%20healthcare,
Retrieved on December 04, 2023.

Kundu, Rohit, (2023). Review on “YOLO: Algorithm for Object Detection Explained
[+Examples]”, Retrieved from https://www.v7labs.com/blog/yolo-object-
detection?fbclid=IwAR26pQ7CBs6xa7qpmmXZtTxXqv-44Dxjbe-AVBqP_t
6hvAeBPwyqEJh-u50, Retrieved on January 17, 2023.

Larkins, Amber, (2021). Review on “Innovation Lifecycles: How to Maximize R&D

ROI Through Collaborative Decision Making”, Retrieved from
https://www.decisionlens.com/blog/innovation-lifecycles?fbclid=IwAR1xkg
AlLFkP0j7jjAeUFsL56OwzSweinpqhXT16VX1rKuuOzKlAvPO31Y#:~:text=
What%20is%20an%20Innovation%20Life,and%20ways%20of%20doing%
20things, Retrieved on September 09, 2021.

Li, Jinling, Hou, Qingshan, Xing, Jinsheng, Ju, Jianguo And, (2020). Review on
“SSD Object Detection Model Based on Multi-Frequency Feature Theory”,
Retrieved from https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=
9078755&fbclid=IwAR2qtGspl3j2C8T9ynLHsdu679Y0po8oNmAXhaS-mn
lvP9ogZ36uMITC0Nw, Retrieved on May 15, 2020.

Pamula, Hanna, Bowater, Jack, (2023). Review on “Propagation Delay

Calculator”, Retrieved from https://www.omnicalculator.com/other/
propagation-delay#how-do-you-calculate-propagation-delay, Retrieved on
June 05, 2023
 73

Shah, Deval, (2022). Review on “Mean Average Precision (mAP) Explained:

Everything You Need to Know”, Retrievedfromhttps://www.v7labs.com/blog
/mean-averageprecision?fbclid=IwAR26pAajrErhc9I29eIjnK53nbIDwG8vQ
4aAA4aIuOZLO063clZ91n_cGoQ, Retrieved on March 07, 2022.

Solawetz, Jacob, (2020). Review on “What is YOLOv5? A Guide for Beginners”,

Retrieved from https://blog.roboflow.com/yolov5-improvements-and-
evaluation/?fbclid=IwAR08lCWDMC3lBe0wV2zWO7yx5is_Zf48k3gRnm9
VPLludub-AMZ9bv8nl9s, Retrieved on June 29, 2020.

Umali, Teresa, (2019). Review on "AI-powered system to improve traffic in the

Philippines", Retrieved from https://opengovasia.com/ai-powered-system-
to-improve-traffic-in-thephilippines/?fbclid=IwAR2CjeEJDjtcuAx9AsA5GFX
r9LWfNXKKBqb9ckKYtZpMOMbflgMmO685es, Retrieved on July 15,
2019.

Valerio, Cristina Joy, (2023). Review on “ARM invests minority stake in Raspberry
Pi”, Retrieved from https://www.yugatech.com/news/arm-invests-minority-
stake-in-raspberry-pi/?fbclid=IwAR3IZnsK7Cpx5ryELQiJ-RDCDLUbEBo-
jp76KYH_GADZ5XkKZx1gk73BqNo, Retrieved on November 04, 2023.

Vatti, Rambabu, (2017). Review on “International Journal of Research Study of Wi-

Fi Signal Strength Measurement and it"s Optimization”, Retrieved from
https://www.researchgate.net/publication/344570581_International_Journa
l_of_Research_Study_of_WiFi_Signal_Strength_Measurement_and_its_O
ptimization?fbclid=IwAR1yBeJUWBNZdLpML375FaYrUkL58ZuqHtc63Wc
ZOV6zLpI_QpR44ETv9Po, Retrieved on December 2017.

Widartha, Vandha, Ra, lIkyeun, Lee, Su-Yeon, Kim, Chang-Soo, (2024). Review
on “Advancing Smart Lighting: A Developmental Approach to Energy
Efficiency through Brightness Adjustment Strategies”, Retrieved from
https://www.mdpi.com/2079-9268/14/1/6, Retrieved on January 15, 2024.

Wu, Cathy, (2021). Review on “Traffic Flow Theory”, Retrieved from

https://web.mit.edu/1.041/www/fall2021/www/lectures/L5-traffic-flowtheory
-2021fa-pre.pdf?fbclid=IwAR0DlzxXxKxiQ29ku3EREXax7OTISCWcj4wm
wndiAB3lQNhWgO-bndRJkYg, Retrieved on September 09, 2021.

Yasay, Jeffrey John, (2021). Review on “Internet of Things in the Philippine

Agribusiness”, Retrieved from https://www.researchgate.net/publication/35
2369025_Internet_of_Things_in_the_Philippine_Agribusiness?fbclid=IwA
R0Mvdg455toNeWmUCQ5QDFddqAi4JeQoHBW-LVEJrRJWd_JXxsj7Ku
ZDPE, Retrieved on June 01, 2021.
 74

APPENDICES
 75

APPENDIX A

GANTT CHART OF ACTIVITIES

For the Object Detection Model Traffic Light System

Research 1
Second Semester, SY 2023-2024
Gantt Chart of Activites

Republic of the Philippines

UNIVERSITY OF RIZAL SYSTEM
Morong, Rizal

OBJECT DETECTION MODEL TRAFFIC LIGHT SYSTEM

First and Second Semester S.Y. 2023 – 2024

Research 2023 2024

Project Plan Aug Sept Oct Nov Dec Jan Feb Mar Apr May
1. Formulation of
5 Research Titles
2. Proposal and
Submission of
Titles
3. Approval of
Title

4. Development
of
Tracker/Timeline
for the Paper and
End Product
5. Consultation
and Guidance
6. Interview of the
Offices
7. Data Collection
from Books,
Journals, Articles,
and Interviews

8. Data Analysis
9. Planning of
System Design
 76

10. Chapter 1
discussion
11. Development
of Background of
the Project
12. Chapter 2
discussion
13. Development
of Design
Methodology
14. Consultation
to Adviser
15. Revision of
Chapter 1 and 2
16. Submission
of Final Draft
17. Colloquium
Oral Defense
18. Revision of
Recommendation
and Suggestions
19. Searching
and Studying of
the required
components
20. Construction
of the Design
21. Development
of the system
22. Testing the
System
23. Survey of
Questionnaire
Checklist
24. Interpretation
of the Gathered
Data
25. Chapter 3
and 4
26. Final Defense

Prepared by:
REYDENTOR L. CASALJAY
 77

VICTORIO DANIEL SD. CRUZ

PETER LUIS P. FENIQUITO
ERON RESTY V. SIOCO

Conformed by:

PAUL ARVY V. ALFONSO, CPE

Research Adviser
 78

APPENDIX B

QUISTIONNAIRE CHECKLIST

Republic of the Philippines

UNIVERSITY OF RIZAL SYSTEM
Morong, Rizal

COLLEGE OF ENGINEERING
Bachelor of Science in Computer Engineering

OBJECT DETECTION MODEL TRAFFIC LIGHT SYSTEM

Part I. Personal Data (Optional)

Name: ______________________________
(Pangalan)

Expert LGU Personnel Community

Part II. Level of Acceptability of an Object Detection Model Traffic Light

System.
Direction: Using the scale below, rate the Object Detection Model Traffic Light
System according to the criteria by ensuring that the relevant column is checked.
(Direksyon: Gamit ang sukat sa ibaba, i-rate ang Object Detection Model Traffic
Light System ayon sa pamantayan sa pamamagitan ng pagtiyak na ang
nauugnay na column ay nasuri.)

Scales are as follows:

5 – Very much Acceptable (Sobrang katanggap-tanggap)
4 – Much Acceptable (Lubhang katanggap-tanggap)
3 – Acceptable (Katanggap-tanggap)
2 – Slightly Acceptable (Bahagyang katanggap-tanggap)
1 – Not Acceptable (Hindi katanggap-tanggap)
 79

1. Functional Suitability 5 4 3 2 1

1.1 The developed system effectively adjusts to variations in

lighting conditions, weather, and other environmental factors.
(Ang binuong sistema ay epektibong umaangkop sa mga
pagkakaiba-iba sa mga kondisyon ng pag-iilaw, panahon, at
iba pang mga salik sa kapaligiran.)
1.2 The developed system consistently achieves accurate
detection of vehicles.
(Patuloy na nakakamit ng binuong sistema ang tumpak na
pagtuklas ng mga sasakyan.)
1.3 The use of the developed system has contributed to a
reduction in road rage incidents in the area.
(Ang paggamit ng binuong sistema ay nag-ambag sa
pagbawas sa mga insidente ng road rage sa lugar.)
1.4 The developed system prevents traffic congestion in the
area.
(Pinipigilan ng binuong sistema ang pagsisikip ng trapiko sa
lugar.)
1.5 The signal timing implemented in the system is efficient.
(Ang timing ng signal na ipinatupad sa system ay mahusay.)
2. Compatibility 5 4 3 2 1

2.1 The developed system is compatible with any USB type

camera.
(Ang binuo na sistema ay katugma sa anumang uri ng USB
camera.)
2.2 The developed system is applicable to other location.
(Ang binuo na sistema ay katugma sa anumang uri ng USB
camera.)
2.3 The developed system is transferable to a new
administrator.
(Ang binuong sistema ay maililipat sa isang bagong
administrator)
2.4 The logic device of the developed system can be
improved to achieve better performance in detection.
(Ang logic device ng binuong sistema ay maaaring mapabuti
upang makamit ang mas mahusay na pagganap sa
pagtuklas)
2.5 The dataset used in the system is upgradable to prevent
inaccurate detection.
(Ang dataset na ginamit sa system ay naa-upgrade para
maiwasan ang hindi tumpak na pagtuklas.)
3. Reliability 5 4 3 2 1
 80

3.1 The developed system consistently provides accurate

results in detecting vehicles.
(Ang binuong sistema ay patuloy na nagbibigay ng mga
tumpak na resulta sa pag-detect ng mga sasakyan.)
3.2 The developed system’s performance is reliable over
extended periods of continuous operation.
(Ang performance ng binuong system ay maaasahan sa mga
pinalawig na panahon ng tuluy-tuloy na operasyon.)
3.3 The developed system reliably detects vehicles under
challenging conditions, such as low visibility or adverse
weather.
(Ang binuo na sistema ay mapagkakatiwalaang nakakakita ng
mga sasakyan sa ilalim ng mapaghamong mga kondisyon,
tulad ng mababang visibility o masamang panahon.)
3.4 The developed system can return to normal operation
when there is a power interruption.
(Ang binuong sistema ay maaaring bumalik sa normal na
operasyon kapag may power interruption.)
3.5 Users can depend on the developed system to
consistently deliver accurate information about traffic light
status.
(Maaaring umasa ang mga user sa binuong system upang
patuloy na maghatid ng tumpak na impormasyon tungkol sa
katayuan ng traffic light.)
4. Usability 5 4 3 2 1

4.1 Users can easily understand and interpret the information

provided by the developed system.
(Ang mga gumagamit ay madaling maunawaan at bigyang-
kahulugan ang impormasyong ibinigay ng binuong sistema.)
4.2 Training and utilizing the dataset require minimal effort
and technical expertise.
(Ang pagsasanay at paggamit ng dataset ay
nangangailangan ng kaunting pagsisikap at teknikal na
kadalubhasaan.)
4.3 Operating the developed system requires minimal effort or
technical expertise.
(Ang pagpapatakbo ng binuo na sistema ay nangangailangan
ng kaunting pagsisikap o teknikal na kadalubhasaan.)
4.4 The website controlling the power of the traffic light is
easy to operate and understand.
(Ang website na kumokontrol sa kapangyarihan ng traffic light
ay madaling patakbuhin at maunawaan.)
4.5 The developed system does its job of guiding the drivers
based on the requirements of the location.
 81

(Ginagawa ng binuong sistema ang trabaho nito sa paggabay

sa mga driver batay sa mga kinakailangan ng lokasyon.)
5. Security 5 4 3 2 1

5.1 The developed system includes strong security measures

to prevent unauthorized access.
(Kasama sa binuong sistema ang mga matibay na hakbang
sa seguridad upang maiwasan ang hindi awtorisadong pag-
access.)

5.2 Sufficient safeguards are implemented to ensure the

privacy of the information handled by the system.
(Ang mga sapat na pananggalang ay ipinatupad upang
matiyak ang pagkapribado ng impormasyong
pinangangasiwaan ng system.)

5.3 The casing of the device is designed to protect the

components inside from attackers.
(Ang casing ng device ay idinisenyo upang protektahan ang
mga bahagi sa loob mula sa mga umaatake)
5.4 The accounts and passwords used in this system are
stored in a safe place.
(Ang mga account at password na ginamit sa system na ito
ay nakaimbak sa isang ligtas na lugar.)
5.5 The communication system used between logic boards
are secured.
(Ang sistema ng komunikasyon na ginagamit sa pagitan ng
mga logic board ay sinigurado)
6. Maintainability 5 4 3 2 1

6.1 The parts of the developed system are affordable.

(Ang mga bahagi ng binuo na sistema ay abot-kaya.)
6.2 The developed system requires less effort to troubleshoot
because it tests itself before operation.
(Ang binuo na sistema ay nangangailangan ng mas kaunting
pagsisikap upang i-troubleshoot dahil sinusubok nito ang sarili
nito bago ang operasyon.)
6.3 Troubleshooting hardware components of the device is
convenient because the parts and wires are properly
organized.
(Ang pag-troubleshoot ng mga bahagi ng hardware ng device
ay maginhawa dahil ang mga bahagi at wire ay maayos na
nakaayos.)
 82

6.4 The parts of the developed system are accessible.

(Ang mga bahagi ng binuong sistema ay naa-access.)
6.5 The developed system is convenient to move.
(Ang binuo na sistema ay maginhawa upang ilipat.)

Comments/Suggestions:

________________________________________________________________

________________________________________________________________

________________________________________________________________

________________________________________________________________
 83

APPENDIX C

CERTIFICATE OF CONTENT VALIDATION

 84

APPENDIX D
 85

LETTER TO CONDUCT THE STUDY

APPENDIX E
 86

LETTER TO THE RESPONDENTS

APPENDIX F
 87

LETTER TO THE PANEL MEMBERS

88
89
90
 91

APPENDIX G

STATISTICAL FORMULA USED

Weighted Mean:

Where:

Mean

X= a row score in a set of scores

Σ= sum of

N= Total number of scores in a set

Confusion Matrix:

Actual

Postive Negative
Confusion Matrix

Predicted Positive True Positive False

Positive

Negative False Negative True

Negative
 92

Jeepney 542 0 0 0 0 0 68

Motorcycle 0 0 0 0 0 0 0

P Tricycle 29 0 2 0 0 0 30
R
Van 0 0 0 177 0 0 24
E
cars 0 0 0 0 190 0 13
D
Truck 9 0 0 0 0 203 32
I
Background 78 1 0 11 5 1 0
C

T Jeep Motor Tricycle Van cars truck Backgrou

ney cycle nd

True

Confusion Matrix of the Object Detection Model:

Propagation Delay:
 93

72 meters
= 0.00000024016 seconds
(2.998 x 10^8 m/s)

Tabular Results of IoU Computation

94

IoU 0.81231 0.89333 0.90445 0.91971 0.83920 0.75180 0.90161

Overlap 150420 281400 1710210 296240 354000 15036 2217905
Area 155595 281400 1841100 310800 375150 15036 2230010
PB
Area 180000 315000 1760000 307541 400680 20000 2447837

Tabular results of Recall and Precision computation

GT
Predicted y2 486 770 1769 795 705 899 1545
Box
(PB) x2 795 790 1490 780 840 690 2115
y1 35 100 154 240 90 815 200
x1 450 370 350 220 230 511 457
Ground y2 800 800 1800 799 745 900 1567
Truth
(GT) x2 800 800 1500 788 876 700 2153
y1 50 100 200 266 115 800 116
x1 400 350 400 211 240 500 466
Class Tricycle Jeepney Truck Cars Van Motorcycle Bus
 95

Classification True False False Recall Precision

Positive Negative Positive

Tricycle 2 0 59 1.0000 0.0327

Jeepney 542 116 68 0.8237 0.8885

Truck 203 2 41 0.9902 0.8320

Cars 190 6 13 0.9694 0.9360

Van 177 11 25 0.9415 0.8762

Motorcycle 0 1 0 0 0

Bus 0 0 0 0 0

Capacity of Object Detection Model Traffic Light System:

Feature Limitation Capacity

Camera Range 4 vehicles

Processing the weights 3220 MS

WiFi-Range 90 meters

APPENDIX H
 96

Summary of Components and Equipment Used

Below is a figure showing a Raspberry Pi 4 that is used in this study to

program the device for the artificial intelligence that is needed to detect the objects

in the camera.

This illustration below shows the node mcu esp8266 that is used to

communicate with the smart traffic light one.

In the study the researchers used a RPI camera as a tool so that the

device can detect objects and also the RPI camera is compatible with the used

microcontroller.
 97

The illustration below shows light bulbs that were used in the smart traffic

light to help vehicles determine to go or to stop.

In this study the Comfast EW73 is used as an access point and provides

connection for the two smart traffic lights so that they can communicate with each

other.

4 channel 5v Relay switch is a device that serves to control the various

components of the traffic signal lights. Each channel typically corresponds to a

specific aspect of the traffic signal, such as red, yellow, and green lights.
 98

The illustration below shows a breadboard power supply that functions as

a power supply for the node mcu.

The Arduino IDE (Integrated Development Environment) is a software platform

used for programming microcontrollers.

 99

Raspberry Pi OS, formerly known as Raspbian, is the official operating

system for the Raspberry Pi single-board computers. It's a Debian-based Linux

distribution optimized for the Raspberry Pi's ARM architecture. Raspberry Pi OS

comes with a range of pre-installed software, including programming tools,

productivity software, and educational resources.

Real VNC is a remote access software that is used to remote or access the

Raspberry Pi OS so that to run the code.

 100

Thonny Python, or simply Thonny, is an Integrated Development Environment

(IDE) for Python programming. It provides features such as code editing,

debugging, and execution of Python code.

PyCharm is an integrated development environment (IDE) specifically

designed for Python development.

 101

APPENDIX I

List and Cost of Materials

Table below present the cost on components and materials used to

construct and successfully come up with the system device with the proper

specification of the project. The table proves that the supplies and materials in the

project construction are readily available in the local electronics and hardware

shop. It also show the total cost of the materials amounting to Php 14,106.2

QTY UNIT DESCRIPTION UNIT TOTAL COST

COST (Php)

1 pcs RPI 4 4GB 3322.20 3322.20

1 pcs PLAIN SHEET 3255.00 3255.00

1 pcs RPI 8MP CAMERA V2 1749.00 1749.00

1 pcs COMFAST CF-EW73 1620.00 1620.00

8 pcs CIRCULAR ACRYLIC 445.00 445.00

GLASS

20 meters FLATCORD 18/2 320.00 320.00

2 pcs RELAY SWITCH 4 318.00 318.00

CHANNEL

2 pcs BOSNY PAINT BLACK 258.00 258.00

1 pcs SANDISK SD CARD 255.00 255.00

64GB

3 packs JUMPER WIRE 195.00 195.00

1 pcs RPI POWER SUPPLY 189.00 189.00

 102

1 pcs NODEMCU V3 ESP8266 160.00 160.00

1 pcs 100 METER BLACK 154.00 154.00

STRANDED WIRE

1 pcs 100 METER RED 154.00 154.00

STRANDED WIRE

500 pcs SCREW 154.00 154.00

1 pcs NODE MCU 120.00 120.00

2 pcs RED COLORED LIGHT 120.00 120.00

BULB

2 pcs GREEN COLORED 120.00 120.00

LIGHT BULB

2 pcs YELLOW COLORED 120.00 120.00

LIGHT BULB

6 pcs SHUTA BULB SOCKET 119.00 119.00

2 pcs 2 GANG UNIVERSAL 118.00 118.00

OUTLET

2 pcs DOOR LOCK 92.00 92.00

1 pcs BREADBOARD POWER 87.00 87.00

SUPLY

2 pcs TERMINAL BLOCK 85.00 85.00

2 pcs C-PLUG WITH CLAMP 78.00 78.00

3 pcs SCREWS WITH NUTS 75.00 75.00

3 pcs HOSE CLAMP 75.00 75.00

 103

2 pcs AB GLUE EPOXY 70.00 70.00

1 pcs RPI CASE 57.00 57.00

780 pcs SHRINKABLE TUBE 52.00 52.00

1 pcs FLAP DISK 50.00 50.00

1.5 KG RIVETS 40.00 40.00

1 pcs DC POWER PLUG JACK 35.00 35.00

1 pcs CUTTING DISK 25.00 25.00

6 pcs WELDING ROD 20.00 20.00

TOTAL AMOUNT OF: Php 14,106.2

 104

APPENDIX J

USER’S MANUAL
105
 106

APPENDIX M

RANDOM PICTURES TAKEN DURING THE CONDUCT OF EVALUATION

 107

PICTURES DURING MAKING THE SYSTEM

 108

THE RESEARCHERS AND THE FINAL SYSTEM

 109

FINAL DEFENSE
 110

CURRICULUM VITAE
111
112
113