Video Analytics for Workplace Safety

The present paper describes how an artificial intelligence-based video analytics system was implemented to improve workplace safety in a construction site in the Middle East. The system can detect and notify potential HSE violations and provide insights and statistics to support HSE management. The system uses high-resolution cameras that stream videos and images to a local server where the video analytics software is installed. The software uses an AI algorithm that can learn to identify unsafe acts and conditions by analyzing the video data and receiving feedback from the HSE operators. The system also sends real-time notifications to the HSE operators via mobile devices. VAWS detects various types of HSE violations, such as PPE noncompliance, work at height, proximity to heavy machinery, and poor housekeeping. The system's reliability, measured by the ratio of confirmed events to total events, increased from 8% to 30% in six months (from November ’23 to April ’24). The system also learned to reduce false positives by receiving specific comments and images from the VAWS analysis team. Thanks also to this solution is possible to identify the riskiest areas and riskiest activities thus allowing site management to provide feedback and specific training to the workers. The implementation of the system faces challenges such as selecting the appropriate type and position of the cameras, complying with explosive zone requirements where necessary, ensuring the network bandwidth and storage capacity, avoiding glare, shadows, and obstructions. The system also requires human validation and intervention, respecting the privacy and ethical rights of the workers, and managing the cyber security risks of the AI system. Based on available results and further planned developments it can be concluded that VAWS is expected to improve the quality and quantity of HSE supervision at large and remote construction sites, especially for routine tasks, night shifts, and less monitored areas.