Project Title: Real-time personnel counting and personal protective equipment (PPE) recognition for construction site safety

Funding: $55,000 (2021-2022)

Project Partners:

		Concordia University – SEISE Lab The Sustainable Energy & Infrastructure Systems Engineering Lab (SEISE), led by Dr. Nasiri, is at the forefront of systems engineering solutions for sustainable energy applications: https://users.encs.concordia.ca/~fuzhan/
	Pomerleau Inc. Pomerleau is a leader in construction industry, originated in Quebec with 4,000 employees across Canada active in Buildings, Civil & Infrastructures Projects: https://pomerleau.ca/
			MITACS Mathematics of Information Technology and Complex Systems: https://www.mitacs.ca/en/programs/accelerate#professor

Project Team

The research partnership team involves Dr. Fuzhan Nasiri (PI) and Dr. Saeed Moradi (postdoctoral fellow) at Concordia University, and Pomerleau team led by Mr. Jean-François Dupuis and Ms. Carolyne Filion.

Project Description

Safety hazards and fatalities in construction sites has always been a main concern for construction and safety managers. Health and safety standards such as the Occupational Safety and Health Administration (OSHA) impose preventive measures by using personal protective equipment (PPE) to reduce the risk of personnel injury. However, using PPE like hardhat, vest, and eye protection are not always respected by the workers due to various reasons. Also, recent situation because of COVID-19 pandemic enforces the companies to comply with governmental regulations regarding the maximum capacity of offices and job sites. Therefore, the main objective of this research is to facilitate safety monitoring of construction sites with real-time control of number of workers and personnel, while automatically identify non-PPE use. In this research project, using computer vision and deep learning algorithms the number of entering and exiting people would be counted to determine occupancy rate. Meanwhile, based on visual data from cameras in different work zones, the workers compliance with the minimum PPE requirements would be monitored using trained deep object detection algorithms. Thereby, any non-PPE use in the job site could be reported instantly to the safety manager as an unsafe condition. In this research, different algorithms will be used to evaluate and compare their performance regarding their accuracy and speed in performing the mentioned tasks. The proposed model is supposed to be a great help for safety and construction managers, while reducing the time and cost of surveillance and decreasing the risks in construction sites.