Construction Site Hazards Identiﬁcation Using Deep Learning and Computer Vision

Oct 22, 2026

Construction sites expose workers to frequent health and safety risks, and despite ongoing attempts to strengthen safety management, accidents still occur and are often detected too late, affecting both worker well-being and project progress. This project develops a computer-vision and deep-learning approach to recognize construction workers, their PPE, and nearby heavy equipment from CCTV footage. Hazard detection is performed from image analysis combined with conditions such as weather, enabling timely contact with onsite safety personnel. A custom dataset trains YOLO-v5 for fast, accurate object detection.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Introduction

Background and safety risks

Contributing factors and preventive measures

Motivation for computer vision and deep learning

Related Work

Prior construction safety algorithms and approaches

sustainability Article Construction Site Hazards Identiﬁcation Using Deep Learning and Computer Vision Muneerah M. Alateeq 1, Fathimathul Rajeena P. P. 1, * and Mona A. S. Ali 1,2, * Citation: Alateeq, M.M.; P. P., F.R.; Ali, M.A.S. Construction Site Hazards Identiﬁcation Using Deep Learning and Computer Vision. Sustainability 2023, 15, 2358. [https://doi.org/](https://doi.org/)[ ](https://doi.org/)[10.3390/su15032358](10.3390/su15032358) Academic Editors: Wei Tong Chen, Ying-Hua Huang and Han-Hsiang Wang Received: 17 December 2022 Revised: 17 January 2023 Accepted: 21 January 2023 Published: 28 January 2023 Copyright: © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . 1 Computer Science Department, College of Computer Science and Information Technology, King Faisal University, Al Ahsa 36291, Saudi Arabia 2 Computer Science Department, Faculty of Computers and Artiﬁcial Intelligence, Benha University, Benha 12311, Egypt * Correspondence: [fatimah.rajeena@kfu.edu.sa](fatimah.rajeena@kfu.edu.sa) (F.R.P.P.); [m.ali@kfu.edu.sa](m.ali@kfu.edu.sa) (M.A.S.A.) Abstract: Workers on construction sites face numerous health and safety risks. Authorities have made numerous attempts to enhance safety management; yet incidents continue to occur, impacting both worker health and the project's forward momentum. To that end, developing strategies to improve construction site safety management is crucial. The goal of this project is to employ computer vision and deep learning methods to create a model that can recognize construction workers, their PPE and the surrounding heavy equipment from CCTV footage. Then, the hazards can be discovered and identiﬁed based on an analysis of the imagery data and other criteria including weather conditions, and the on-site safety ofﬁcer can be contacted. Our own dataset was used to train the You Only Look Once model, version 5 (YOLO-v5), which was put to use as an object detection model. The detection model's performance in tests showed promise for fast and accurate object recognition in the ﬁeld. Keywords: object detection; PPE; heavy equipment; YOLO-5 1. Introduction In the past decade, the worldwide construction sector has experienced explosive growth, and safety management on building sites has become a top priority. However, accidents continue to occur in these locations and are frequently undetected or noticed extremely late. According to the Saudi Arabian Health and Safety Association, commercial and public sector workplace injuries in 2016 totaled over 67,000 and related medical treatment cost over SAR 370 million [1] . Minor and signiﬁcant injuries on the job have ramiﬁcations for the workers, their families and the project's schedule and budget. Therefore, there have been many efforts in recent years to make construction sites safe, productive and smart. Construction sites are known to be highly hazardous environments due to their dynamic and temporary nature. Some studies showed that accidents at construction sites can be caused by many factors, as follows [2,3]: lack of awareness and experience among workers, lack of safety training among workers, workers not wearing personal protective equipment (PPE), no safety ofﬁcer located in danger zones and machinery defects and errors. Construction workers are involved in many activities that might expose them to risksand accidents. Some of the hazards on construction sites may involve falling from heights, being electrocuted, being struck by heavy equipment or falling materials, being caught in equipment and many more, depending on the nature of the site [3] . Those accidents can be prevented by following safety policies and rule

FAQs

What problem does the study address on construction sites?

Construction sites involve many hazards and accidents that may go undetected or be noticed very late, impacting worker health and project momentum.

How is the proposed system intended to identify hazards?

It uses computer vision and deep learning to recognize construction workers, PPE, and surrounding heavy equipment from CCTV footage, then discovers and identifies hazards through image analysis and additional criteria.

Which model and dataset are used for training, and what were the results?

The model is YOLO-v5 trained on the authors’ own dataset, and tests indicated promising performance for fast and accurate object recognition in the field.

Oct 22, 2026

Uploaded by Quinn Holloway