80
RESEARCH PAPER
VEHICLE OVERSPEED DETECTION IN
BLINDSPOTS USING OCR
Dr. M. Manickam, Akshat Kumar, Rokkam Pranathi
SRM Institute of Science and Technology, Chennai, India
Email: {ak5009, rp3484, manickam} @srmist.edu.in
*corresponding author
Abstract - Traffic accidents are one of the traumatic events that cause great loss. In India it has become a big problem because it
has taken the lives of innocent people. The fight against traffic accidents has become an important task. One of the reasons
behind the project is to prevent the increasing number of road accidents caused by speeding in prone areas. This study looks at
how we detect speeding vehicles and their license plates. Vehicle speed detection and tracking contributes significantly to the
safety of citizens by preventing a large number of accidents. This module is essential in traffic monitoring where traffic
efficiency and people's safety are the main concerns of. In this study, we discuss possible ways to find cars, track speed and
track license plates. Many studies and research papers on speed detection have been done before and can be found on the
internet. The recommended approach broadly includes the following steps:
(i) Sketch the license plate number, (ii) Look up the date and time, (iii) Find the location, (iv) Calculate the speed, (v) Report to
the authorities. The project aims to monitor the speed of vehicles to prevent accidents and reduce impacts in the event of an
accident. This project proposes a modern and efficient method to detect and report illegal speed.
Keywords—Contouring, Geo Tag, authorisation, Sobel edge detection, noise removal, Gradient extraction,
pixel, TSC (key words)
I. INTRODUCTION
Our daily activities cannot be separated from transportation, which plays a crucial role. The response of the
government shouldn't be constrained when the need for transit increases in urban areas. just for creating new
infrastructure or expanding existing routes. Over time, it experiences a number of grave problems, such as traffic
accidents and air pollution. In order to maintain a balance between social needs and available transportation
resources, the transportation system must develop.Traffic speed control (TSC) makes this development possible.
Traffic-related concerns have been handled more effectively during the last three years using TSC services. The
integration of cutting-edge technology and local transportation infrastructure is one of TSC's strengths.
This benefit fortifies them, enabling them to be used to monitor traffic flow, spot violations, and increase
driving. Because of this, over the past 10 years, there has been a lot of academic interest in the use of video- based
systems, which has accelerated up the development of traffic planning and engineering applications.TSC can be
used for a variety of tasks, such as categorising vehicles,calculating traffic density, and measuring vehicle speed. In
all of these applications, it is vehicle-based speed measurement that aids in the detection of violations. Cameras
have evolved into a crucial part of the system due to their cheap maintenance needs, wide fields of view, and
improved documentary proof. Computer vision-based techniques are used to process the recorded images from the
cameras.
The rate of movement of a moving object at a specific time is referred to as speed. The time elapsed between the
primary parameters and the distance travelled are used to calculate it. Average speed and instantaneous speed are
the two types of speed. The average speed of the car indicates how far it travelled throughout each interval. And the
instantaneous speed immediately displays the correct speed. The specific pace at which the vehicle is moving at
any given time is referred to as "speed" in the proposed work. The two main techniques used in vehicle tracking
applications are feature-based techniques and motion-basedtechniques.
81
Using feature-based algorithms, each image's features, such as lights, licence plates, wheels, etc., are detected.
This method has a high processing cost and decent accuracy, whereas motion-based methods consider all moving
objects. Due to its consistent qualities and high likelihood of filtering out non-vehicle items, it is also believed to be
distinctive among work vehicle objects. The process is cheap to compute and unaffected by attributes. Motion-
based systems use filtering techniques to take non-vehicular features out of focus zones to make up for this lack of
capabilities. More than 80% of road accidents in India are caused by speeding and negligent driving, and other
people's reckless driving hurts many innocent people.
We have a system in place to automatically identify people who are driving faster than the posted maximum
speed limit on their specific roads or highways since it is imperative that this be checked. The speed detection,
image acquisition, image transmission, and image processing are the focus of this research. The camera can
determine if the posted speed restriction is higher than the posted top speed limit. High-definition cameras
complete the image transmission and acquisition process. To digitise the licence plate characters and separate the
licence plate, the server runs an image processing programme. The number plate number will be submitted to the
authorities after exceeding the speed limit.
Currently, the nation as a whole is using more vehicles. A distinctive vehicle identification number serves as each
of these cars' principal means of identification. The licence number, which denotes the legal right to engage in
public sports, contains the identity. Every vehicle in the world needs a unique licence plate that is permanently
attached to the vehicle's body (at the very least at the back). As the number of vehicles grows, they must recognise
parallel vehicles. This recognition method aids in speed detection and highway safety.
II. LITERATURE SURVEY
The main goal of using a vehicle detection system, which can distinguish non-vehicle things including people,
animals and road surfaces from moving vehicles. The identified vehicles are then observed for various purposes,
including traffic analysis, vehicle identification, and speed measurement. Proposed vehicle speed monitoring in a
fully automated traffic monitoring system for tracking moving objects. Oversight mechanism by the Transport
Commissioner and the National Transportation Administration (TNSTA). Consider the advantages The efficiency
of the system is mainly based on the accurate identification of the lane during self- inspection, detection and
monitoring of vehicles. A 3D vehicle detection method based on probabilistic assumptions for difficult footage is
proposed. Here, the width and length of the moving vehicle are calculated using the background subtraction
method. Finally, the third dimension of the bubble object is calculated using the homography matrix. The
probability of misclassifying the vehicle type is the main drawback of this strategy due to its size. [1]
The image quality of giant cars can be used to distinguish them from small cars, the contours and margins of small
cars. Another system that detects cars in heavy traffic situations uses the AOG architecture. Here, the windbreak
serves as a monitoring indicator. mainly in covert traffic situations. The Active-Pace model and the windshield are
divided into different sections for each station shown on the map. A method of bandwidth conversion matching the
vehicle's input parameters used by the vehicle's terminal. In latency analysis, the vehicle type is determined,
classified and statistically the number of cars visible in the video. This method has the advantage of using less
computer power because it monitors vehicles in groups in heavy traffic conditions. YOLO V3 Object Detection
Algorithm To achieve extremely accurate parking detection, a deep convolutional neural network is used.
Clustering algorithm solves the media routing problem[2].
82
Based on the position of the lights and number plates, a new detection system is provided for vehicles of different
colors and configurations. A sophisticated Kalman filtering method is then used to monitor the identified vehicles
on low frame rate footage. In relative discrimination, directional slope histograms are also used. A proposed low-
light speed measurement method is detected and fixed on the vehicle's headlights. The basic problem of this system
is the integration of the headlights of some cars, which is accomplished by the cross-contact technique and the
pinhole method. Elsaq technology for the telescopic system was used to create a license plate reading (LPR)
simulation model for speed measurement. and monitoring by specifying the position, height and angle of the
camera. In a stereo vision system based on license plate detection, a measurement method that calculates vehicle
speed and its associated time is introduced. License plate characters are recognized in this system by OCR and
CNN methods. The basic goal of all the strategies discussed in the literature review is to create a system that can
operate nearly as efficiently when there is a disruption in the flow of traffic. The proposed system attempts to
overcome this difficulty by marking an area of interest on each vehicle to be provided. [3]
III. PROPOSED METHOD
The technique described in this article uses the video output of a fixed camera to observe the driving characteristics
of the vehicle in order to calculate instantaneous speed. The strategy focuses on urban areas due to slowing car
speeds. The proposed method does not depend on the physical properties of the vehicle, such as license plates or
front/rear lights, thus reducing computation time as it does not require matching of the displayed characteristics.
The identified vehicle is further tested on successive frames to determine the corresponding speed. By counting the
number of frames required for the vehicle to move through a particular position, this can be determined. The five
distinct steps of the proposed technique are illustrated in Figure 1. The system makes several assumptions, such as
mounting a camera at a specified height, which is used to measure vehicle speed as well as light and road surface.
The five stages used in the recommendation system will be explained in more detail in the rest of this article.
83
A. Preprocessing
Initially, two cameras were installed. When an image is captured in a camera, the image data is extracted including
date, time and location coordinates. The data needed for speed calculations is then collected and calculated. If the
calculated speed is normal, the process has ended. If not it would carry onwith the other steps and It will alert the
authorities if the speed is exceeded. These cameras for better accuracy will instead of just capturing picture would
capture the video of the vehicle with date and time stamp so that the number plate detection can be a lot more
easier and a lot more accurate incase of a blurpicture because of the weather. Once the clear image of the number
plate has been taken and from camera 1 and camera 2 and over-speed has been detected then the number plate
would be contoured for number plate detection to inform the authorities . The authorities being traffic police and
the car company for the warning.
B. Contouring
• Contouring is the process by which software takes an image of an input vehicle. Then, once it has captured the
license plate, it will place the pointer over the edge of the clamped license plate. Then the image is cropped.
• Due to different viewing angles in different frames, the overall shape of the moving vehicle is made up of
several distinct components in the final binary image. Using a method, such as optical flow, each area of
motion is represented by a motion vector, and a set of vectors is then aggregated into a unit through homogeneity
analysis. In the proposed strategy, objects are presented as vehicles for detection by changing theimage.
• These increments are performed with operations involving the binary image and the kernel image, for example,
as two inputs. Use open and close functions between different image transformations. Due to its brightness and
proximity, the Vulgate system fills the object's additional elements with predefined pixels.
• Sketch requires the Sobel operator to extract edge properties. It is capable of generating different x and y
gradients. The approximate slope of the grayscale image is determined by a separate differential operator. The
likelihood of it being an edge increases with the size of the gradient. Gaussian differentiation and smoothing are
guaranteed by the operation of the Sobel operator.
• Sobel slope is determined as follows:
Convolution I with a one-dimensional kernel Gx, varying degrees. For example, Gx is defined as follows when the
kernel size is 3:
• Vertical variation:
Rotate I with a one-dimensional kernel Gy. For example, the Gy calculation for a kernel size of 3 is:
84
• To determine the approximate gradient, combine the two examples above in each pixel of the image:
G = Gx + Gy. The duplex image is after discovering the morphological image generated from the Sobel edge. After
morphological image processing, further continuous processing is performed due to roughness of edges of some
images. In other words, edge detection of image Curve connecting all adjacent points (with contour) with same
color or intensity is called image edge. Plates are recognized when the edges and curves of the image are joined
together.
C. Location detection
Geotagging is used to find the location of the camera based on the latitude and longitude information provided as
input to the system. The generated CSV file will contain the exact location of the vehicle along with the date,
providing the details we need for quick calculations. Geotags allow you to locate the camera that took the picture.
This makes it easy to determine where the velocity is happening. Geolocation is the process of linking geographic
coordinates to a mobile phone's location-based advertising. Geotagging can be used for images, videos, web pages,
text messages, and QR codes, and can also include timelines or other content information. Organizations can
deliver high-quality services and messages with their brand. Geotags can also tell you where someone is when they
visit a website or where they move during the day using their mobile device.
Geotagged photos contain location information in the image file. Most phones have an internet connection and
the camera automatically disables GPS tracking. They also record the date and time the photo was taken. Real-time
geolocation refers to automated processes for collecting information (such as images, audio or video), associating a
specific location with information, publishing information online and publish information in a timely manner. As
such, it is an extension of automatic geolocation that must be performed or tied to location, but also requires
communication with wireless data transmission systems such as local area networks.
LAN connection is required to transfer data from camera 1 and camera 2 inputs to our vehicle speeding and license
plate detection project.
D. Reporting to the authorities
When a vehicle is detected speeding, an RTO fine will be sent to the vehicle owner, and a notification will be sent
to the car manufacturer to warn the vehicle owner and notify them of the consequences.
IV. RESULT
In the results, the demonstration shows examples of recommended methods based on detection speed and accuracy.
Some vehicle images are included in the computer's video input and output data, in addition to the relative speed
and time at which the image was received by the computer. Additional information such as frame counters and
media IDs are stored in the CSV file.
85
A. Dataset
Two datasets in image format are used to evaluate the performance of the proposed system. These images are first
contoured and the image information is collected in a csv file. We added the option to record video to get the
clearest image and then use it to detect more license plates by contour.
B. Vehicle Detection
The vehicle is detected in the video recording camera to get the most accurate number plate image. After obtaining
the clearest image, the image will be further contoured to detect license plates. By using pytessaract and geotag,
details such as vehicle number, date, time and image location are detected and collected in a csv file. After the
license plate is detected, the fine will be sent to the vehicle owner through E-challan, which can be paid online.
C. Speed Detection
According to the output of camera 1 and camera 2 the speed of the vehicle is found out and if it violates the
maximum speeding it is reported to the authorities. The speed can be found out by the data such as time noted by
both cameras and distance travelled, the speed of the vehicle can be detected. The speed is detected using the
formula:
𝑑
θ=
𝑡2 − 𝑡1
where 𝜃 --→ speed calculated
d --→distance travelled by the vehiclet1 --
→time stamp noted by the first camera
t2 --→time stamp noted by the second
camera
V. CONCLUSION
In conclusion, we provide the necessary framework to analyze vehicle speed and license plate. We also talk about
speed detection using different cameras (at least 2) and understand how they work and the data involved. We also
include screenshots from Camera 1 and Camera 2 in the dataset obtained from recorded video to get the clearest
image for accurate results. To parse objects we use opencv. Safety and risk reduction are key issues that need to be
prioritized. The rules must be followed to continue the tutorial fast enough. Technology and monitoring devices
that monitor vehicle movement and speed as well as license plate detection can help reduce the frequency of road
accidents and track incidents, especially on the highways. . This project/documentary focuses on the violation of
traffic rules and the appropriate measures to be taken to reduce traffic accidents thathave claimed many lives.
REFERENCES
[1] Anhang Wang, “Vehicle recognition algorithm based on improved YOLOV3”, 2020 IEEE 9th Joint
International Information Technology and Artificial Intelligence Conference (ITAIC), Chongqing.
86
[2] Sushant Kadam, Rushikesh Hirve, Nikhil Kawle, Payal Shah, “Automatic Detection of Bikers with No Helmet
and Number Plate Detection”, 2021 12th International Conference on Computing Communication and
Networking Technologies (ICCCNT), Kharangpur, India.
[3] B.KrishnaKumar, K.Kousalya, R.S.Mohana, E.K.Vellingiriraj, K.Maniprasanth, E.KrishnaKumar “Detection
of Vehicle Speeding Violation using Video Processing Techniques”, 2022 International Conference on
Computer Communication and Informatics (ICCCI), Coimbatore,India.
[4] Kaviya , Manasa “AUTOMATIC DETECTION OF OVERSPEED VEHICLE ” , IEEE 5th may 2022
[5] Shraddha S. Ghadage , Sagar R. Khedkar, “3Automatic Number Plate Recognition System Using Machine
Learning Algorithms” , International Journal of Engineering Research & Technology (IJERT) ISSN: 2278-
0181 , 6th January 2020.
[6] Pradeep Kanavi, Chaithra K B, Chaitra K T, Bhoomika M G, Lakshmi C T,” Automatic Vehicle Over speed
Detection Alert and Controlling system on highway ”, International Journal of Engineering Research &
Technology (IJERT), 30th October 2022.
[7] Rishabh Rana, Ms. Nidhi Sengar, Ms. Vasudha Bahl, Dr. Amita Goel,” Number plate detection ofover
speeding Vehicles” International Journal of Engineering Research & Technology (IJERT), 12th December
2021.
[8] N.Palanivel Ap; T. Vigneshwaran; M.Sriv Arappradhan; R. Madhanraj“Automatic Number Plate Detection in
Vehicles using Faster R-CNN”IEEE-2020 International Conference on System, Computation, Automation and
Networking (ICSCAN),30 November 2020.
[9] Pratik Sharma; Shivam Gupta; Pritesh Singh; Kunal Shejul; Dinesh Reddy “Automatic Number Plate
Recognition and Parking Management” 2022 International Conference on Advances in Computing,
Communication and Applied Informatics (ACCAI), 15 April 2022.
[10] Vijaya Vardan Reddy SP; B. Sathyasri; A. Balaji; S. Vanaja; Rahul Krishnan “Automatic NumberPlate
Recognition System for Entry and Exit Management” 2021 6th International Conference on Communication
and Electronics Systems (ICCES) IEEE - xplore, 02 August 2021.
[11] B.KrishnaKumar; K.Kousalya; R.S.Mohana; E.K.Vellingiriraj; K.Maniprasanth; E.KrishnaKumar.,
“Detection of Vehicle Speeding Violation using Video Processing Techniques ” , IEEE 31st Feburary 2022.
[12] Mohammed GufranHaseeb, Ankit Kumar , Sujeet Kumar Horo, Ashish Tiwary “Highway Overspeed
Detecter and Alarming System”, International Journal of Engineering Science and Computing,March 2021.
