P.S.

V COLLEGEOFENGINEERINGANDTECHNOLOGY
(An Autonomous Institution)
(Approved by AICTE,New Delhi and Affiliated to Anna University, Chennai)
Accredited by the NAAC with ‘A’ grade
(Inclusion Under Section 2(f) & 12(B) of the UGC Act, 1956)
(AnISO9001:2015CertifiedInstitution)
Mittapalli, Balinayanapalli Post, Krishnagiri – 635 108

A PROJECT REPORT

Submitted By

DHARANEESWARAN .S
611823205015
KAVASKAR .M
611823205032
MANIKANDAN .J
611823205036

BACHELOR OF TECHNOLOGY
In
INFORMATION TECHNOLOGY

P.S.V COLLEGEOFENGINEERINGANDTECHNOLOGY
(An Autonomous Institution)
(An ISO 9001:2015 Certified Institution)
KRISHANAGIRI -635108
 PROJECT TITLE
CONSTRUCTION SITE SAFETY
ENSUREMENT

ABSTRACT
Construction sites are inherently
hazardous environments where
worker safety is often compromised
due to human error, lack of
monitoring, or delayed emergency
response. The increasing rate of
injuries and fatalities at such sites
underscores the urgent need for
advanced safety solutions. This
project introduces an AI-powered
Construction Site Safety Ensurement
System that utilizes computer vision,
machine learning, and real-time alert
systems to monitor worker safety
and prevent accidents.
At its core, the system leverages
deep learning-based object detection
algorithms (such as YOLOv8) to
ensure that all personnel on-site
comply with personal protective
equipment (PPE) standards —
specifically the use of helmets and
safety vests. Any deviation from
safety protocols triggers immediate
alerts to supervisors through email
and dashboard notifications.
Beyond safety compliance, the
system is designed to detect critical
accidents such as workers falling,
sudden collapses, or the presence of
unauthorized personnel in danger
zones. These events are detected
using
AI-based activity recognition models
and posture analysis techniques.
When a high-risk event occurs, the
system not only logs the incident but
also takes automated actions such
as contacting emergency services
(ambulance, fire department) and
notifying site engineers and
managers via email or SMS.
The platform also includes real-time
danger zone detection, where
hazardous areas are dynamically
marked using AI to help workers
avoid accident-prone zones. It
provides an interactive dashboard
for site managers to view alerts,
analyze safety patterns, and review
historical incidents, improving long-
term safety planning.
In addition to visual data, the system
can integrate with environmental
sensors (IoT devices) to detect
smoke, gas leaks, or abnormal
temperature conditions, further
enhancing situational awareness.
By integrating cutting-edge
technologies and automation, this
project delivers a comprehensive,
scalable, and proactive safety
management solution for
construction sites. It minimizes
human oversight, shortens response
times during emergencies, and
fosters a culture of safety and
compliance in high-risk
environments. Ultimately, it
contributes to reducing workplace
injuries, saving lives, and improving
operational efficiency in the
construction industry.

Key Functionalities:
Object Detection: Real-time
detection and tracking of safety
equipment such as helmets, safety
vests, and construction machinery
using deep learning-based models
like YOLOv8 or SSD.
Worker Safety Ensurement: AI
models continuously verify whether
workers are wearing mandatory
protective gear. Non-compliance
immediately triggers an alert
system.
Accident Detection: The system
identifies potential worksite incidents
including:
Falling workers
Site collapses
Unauthorized entries into danger
zones
Danger Zone Detection: The
environment is mapped and marked
with dynamic danger zones, where AI
detects risky activities or unsafe
behaviors, prompting warnings.
Email Alerts & Emergency
Notifications:
Automatic email alerts to site
engineers and managers for non-
compliance or hazardous events.
Emergency prevention
integration with:
Ambulance calling systems
Fire department alerting systems
These are activated automatically
based on severity (e.g., worker
collapse, fire, structural damage).
AI-powered Environment
Monitoring: The system integrates
environmental sensors and visual
feeds to detect:
Unsafe temperature/humidity levels
Smoke/fire
Slips or structural weaknesses
Technologies Used: Computer
Vision: OpenCV, YOLOv8, MediaPipe,
TensorFlow/PyTorch

Backend & Alerts: Python

(Flask/Django), SMTP for email,
Twilio/Fire service APIs for call alerts
Frontend Dashboard: HTML, CSS,
JS, Bootstrap, Charts.js for
visualization
Database: SQLite/MySQL for logs,
alerts, worker profiles, and safety
records
Edge Devices/Cameras: CCTV/IoT
camera feeds integrated for real-
time analysis
Conclusion:
This project delivers a
comprehensive AI-based safety
management system for
construction sites. It not only
enhances worker protection through
real-time PPE detection but also
plays a pivotal role in accident
prevention, emergency response,
and site hazard mitigation. Through
intelligent alerts and environmental
awareness, it ensures a proactive
approach to construction safety,
reducing fatalities and improving
compliance across large-scale
construction operations.
History of Accidents in
Construction Sites:
Overview:
Construction is one of the most
dangerous industries in the world,
accounting for a significant number
of occupational injuries and
fatalities. Despite technological
advances and the implementation of
safety regulations, construction sites
continue to pose serious risks due to
factors like human error, lack of
supervision, and unsafe working
conditions.
Global Statistics:
According to the International Labour
Organization (ILO), the construction
industry accounts for approximately
30% of all fatal workplace accidents
globally.
The U.S. Bureau of Labor Statistics
(BLS) reported that in 2023, the
construction industry had the
highest number of fatal work injuries
among all industries, totaling over
1,000 deaths.
The Occupational Safety and Health
Administration (OSHA) in the U.S.
highlights the “Fatal Four” causes of
construction worker deaths:
Falls – 34%
Struck by object – 11%
Electrocution – 8%
Caught-in/between – 5%
These four causes alone are
responsible for nearly 60% of
construction-related deaths.
Historical Examples of Major
Construction Accidents:
1. Mecca Crane Collapse (Saudi
Arabia, 2015):
A large crawler crane collapsed at
the Grand Mosque in Mecca, killing
111 people and injuring over 390.
Cause: Unsafe crane placement,
strong winds, and lack of hazard
mitigation.

2. Hard Rock Hotel Collapse (USA,

2019):
A hotel under construction in New
Orleans collapsed, resulting in 3
deaths and over 18 injuries.
Cause: Structural failure, poor
construction planning.
3. Delhi Building Collapse (India,
2020):
An under-construction building in
Delhi collapsed, killing 3 workers and
injuring others.
Cause: Overloading of materials,
weak structure, and lack of safety
compliance.
4. Mumbai Building Collapse (India,
2021):
A residential construction site
collapsed due to poor quality
materials and rainwater damage.
Death toll: 11 people, most of them
construction workers or residents.
Why AI-based Safety Systems
are Necessary:
Traditional safety methods depend
on human monitoring, which is prone
to fatigue, distraction, and slow
reaction times.
AI systems can analyze thousands of
frames per minute, detect anomalies
instantly, and alert responders in
real-time, saving lives.
Proactive detection of unsafe
behaviors (like not wearing helmets)
can prevent accidents before they
occur.
Literature Survey
The integration of artificial
intelligence (AI) and computer vision
in safety systems has attracted
increasing attention in recent years,
particularly in high-risk sectors like
construction. Various studies and
technologies have emerged to
monitor safety compliance, detect
accidents, and automate emergency
responses. This section reviews the
significant works and research
efforts relevant to this project.
1. Safety Equipment Detection Using
Deep Learning
Reference: “Real-Time Detection of
Safety Helmet Wearing on
Construction Sites Based on Deep
Learning”
Authors: Zhang et al., Sensors
Journal (2020)

Summary:This study proposed a

YOLOv3-based object detection
method to identify whether workers
are wearing helmets. The model
achieved high accuracy in real-time
detection and was deployed in live
construction environments.
Relevance: Demonstrates how deep
learning can be used for real-time
PPE (helmet) compliance monitoring.
2. Behavior Analysis for Accident
Detection
Reference: “Human Fall Detection
from Surveillance Video Using Deep
Learning”
Authors: Cheng and Wang, IEEE
Transactions on Pattern Analysis
(2019)
Summary:
This paper presented a fall detection
system using convolutional neural
networks (CNNs) and long short-term
memory (LSTM) networks to
understand posture and motion
patterns.
Relevance: Provides a foundation for
fall detection systems using posture
and motion tracking.
3. AI in Construction Safety
Management
Reference: “Application of Artificial
Intelligence in Construction Safety
Risk Assessment: A Review”
Authors: Hwang et al., Automation in
Construction (2021)
Summary:
The research reviewed various AI
applications in safety risk
assessment, including video
analytics, sensor-based monitoring,
and data-driven hazard prediction.
Relevance: Highlights the growing
trend of AI integration for safety
forecasting and automation in
construction.
4. Emergency Alert Systems Using
IoT
Reference: “IoT-Based Emergency
Alerting and Health Monitoring
System for Construction Sites”
Authors: Kumar and Bhatia,
International Journal of Emerging
Technology (2020)

Summary:
This study implemented a
microcontroller-based IoT alert
system that could monitor heart
rate, temperature, and fall detection,
triggering SMS/email alerts.
Relevance: Supports the integration
of sensor-based emergency
detection and automated
communication.
5. Danger Zone Recognition Using
Computer Vision
Reference: “Hazardous Area
Detection and Visual Warning
System Using AI”
Authors: Lin et al., IEEE Access
(2021)
Summary:
The paper introduced a visual
warning system that identifies high-
risk zones using surveillance video
and object motion prediction.
Relevance: Shows how AI can
dynamically detect and alert workers
in unsafe areas.

6. Helmet and Vest Detection Using

YOLOv4/YOLOv5
Reference: "Safety Gear Detection
for Construction Workers Using
YOLOv5"
Authors: Mishra et al., Elsevier
Procedia Computer Science (2022)
Summary:
A custom-trained YOLOv5 model was
used to detect helmets and vests
with an accuracy of 95%,
demonstrating the model's
effectiveness in real-time
applications.
Relevance: Confirms the reliability of
modern YOLO models for detecting
multiple safety gear items.
SCREENSHORTS: