Illegal parking [9] in cities is a critical problem in urban traffic management, pedestrian safety, and visual appeal. Conventional mechanisms are reliant on the visual monitoring capabilities of law enforcement officers and citizen complaints that are expensive and time-delayed. In the recent past, the availability of public camera infrastructure and research in object detection using deep learning enabled the development of real-time, automated mechanisms for enforcing parking regulations. In this paper, a novel methodology is proposed that utilizes publicly available camera feeds, the most advanced object detection models, and a spatial analysis based on polygon detection to identify and flag illegally parked vehicles. By modeling restricted zones as polygons within the camera’s field of view and by integrating a temporal persistence criterion, the approach in this paper correctly identifies vehicles that remain stationary in no-parking areas beyond some pre-defined threshold, discarding false positives along the way. It will be shown that the pipeline is scalable and robust for large-scale deployments through continuous video tracking and YOLO-based detection. It indicates a promising direction toward smart city initiatives that may enable automated and proactive detection of traffic violations.

Computer Vision Traffic Surveillance YOLO Illegal parking detection YOLO object detection Multi-object tracking Public CCTV monitoring Polygon-based spatial analysis Automated traffic enforcement Temporal violation criteria Real-time video analytics