Detection of Personal Protective Equipment (PPE) Compliance on Construction Site Using Computer Vision Based Deep Learning Techniques

Abstract

Construction safety is a matter of great concern for practitioners and researchers worldwide. Even after risk assessments have been conducted and adequate controls have been implemented, workers are still subject to safety hazards in construction work environments. The need for personal protective equipment (PPE) is important in this context. Automatic and real-time detection of the non-compliance of workers in using PPE is an important concern. Developments in the field of computer vision and data analytics, especially using deep learning algorithms have the potential to address this challenge in construction. This study developed a framework to sense in real-time, the safety compliance of construction workers with respect to PPE, which is intended to be integrated into the safety workflow of an organization. The study makes use of the Convolutional Neural Networks model, which was developed by applying transfer learning to a base version of the YOLOv3 deep learning network. Taking into account the presence of hardhat and safety jackets, the model predicts compliance in four categories such as NOT SAFE, SAFE, NoHardHat, and NoJacket. A data set of 2,509 images was collected from video recordings from several construction sites and this web-based collection was used to train the model. The model reported an F1 score of 0.96 with an average precision and recall rate at 96% on the test data set. Once a non "SAFE" category is detected by the model, an alarm and a timestamped report are also incorporated to enable a real-time integration and adoption on the construction sites. Overall, the study provides evidence on the feasibility and utility of computer vision-based techniques in automating the safety-related compliance processes at construction sites.

PPE detector: a YOLO-based architecture to detect personal protective equipment (PPE) for construction sites

“…The YOLOv3 architecture was used by Delhi, Sankarlal & Thomas (2020) to predict four classes such as NOT SAFE, SAFE, NoHardHat, and NoJacket. The authors trained their model using 2,509 images that were collected from video recordings from the construction sites and internet-based collections.…”

Temporal Visual Patterns of Construction Hazard Recognition Strategies

“…(RBC). Then, the presence of hidden hazards can be judged by comparing human posture and wear with safety norms (RBC + TM), for example, identifying unsafe behaviors and abnormal postures based on skeletal movements [19,20] and determining whether workers are wearing personal protective equipment (PPE) such as helmets and safety belts [7,21,22]. Even though such algorithms are computationally efficient and reliable, most of them discriminate low visual complexity hazards with a fixed perspective, which is difficult to adapt to complex and dynamic construction scenarios.…”

A Novel Implementation of an AI-Based Smart Construction Safety Inspection Protocol in the UAE

“…A recent paper titled, "Detection of Personal Protective Equipment (PPE) Compliance on Construction Site Using Computer Vision Based Deep Learning Techniques" developed a framework to detect in real-time, the safety compliance of construction workers with respect to PPE [16]. YOLOv3 was their adopted deep learning network that was used in their research, and it proved its ability to perform in a construction site environment with a relatively high accuracy.…”

“…YOLOv3 has proved to be effective at detecting multiple objects at construction sites [16]. Thus, YOLOv3 was chosen as the primary CNN architecture for this research.…”

“…The results obtained in this research were as high and accurate as those results obtained in the earlier studies. For instance, according to [16] a similar YOLO v3 model were trained to detect the availability of safety helmets and safety vests in a construction site. Their model achieved an overall precision of 96%.…”