Online Safety Zone Estimation and Violation Detection for Nonstationary Objects in Workplaces

This study presents a deep neural network (DNN)-based safety monitoring method. Nonstationary objects such as moving workers, heavy equipment, and pallets were detected, their trajectories tracked. Time-varying zones (SZs) of estimated based on trajectories, velocities, proceeding directions, formations. SZ violations are defined by set operations with sets points in the SZs object trajectories. The proposed methods tested using images acquired CCTV cameras virtual 3D simulations plants loading docks. DNN-based detection tracking provided accurate online estimation time-varying that adequate for workplace. operation-based violation definitions flexible enough to monitor various scenarios currently monitored workplaces. can be incorporated into existing site systems single-view at vantage points.