Real-Time Human Tracking and Predictive Danger Zone Forecasting for Public Safety

Community safety and security in high-density urban environments such as public transport terminals, commercial zones, schools, colleges, and places of public assembly need precautionary monitoring of pedestrian movement to mitigate accidents, risks, and hazards. Existing and available surveillance infrastructure primarily supports incident-based monitoring with limited prediction-driven capability to identify risks and hazards prior to occurrence. This proposed research paper introduces a novel implementation framework titled "Real-Time Human Tracking and Predictive Danger Zone Forecasting for Public Safety". The proposed system processes and analyzes video surveillance recordings captured by the existing CCTV cameras, reduces the resolution from 720p to 480p using gray-scale conversion to reduce computational complexity, applies YOLO variants based on human object detection, and utilizes light-weight AI-powered tracking to investigate the trajectories of persons in the crowd environment. Explicitly defined high-risk and hazard zones are marked, and the proposed system focuses on single or multiple persons entering from multiple trajectories who may move towards danger or hazard-prone areas. Real-time alerts are produced as output to support early-stage intervention to reduce risks and hazards. Primary analysis demonstrates that the proposed method delivers high-performance, accurate human detection and predictive accuracy without using GPU accelerations, while maintaining computational viability for real-time visualization.