Parking Violation Detection using Computer Vision for Urban Traffic Management

Parking violations disrupt traffic flow and accessibility, and most often these violations require resource-intensive manual enforcement. This paper examines the real-time parking violation detection system-Supervised Automated Monitoring-System. This employs the YOLOv5 state-of-the-art object detection model, which facilitates the accurate detection of vehicles that are prohibited from parking. The experimental results demonstrate greater than 95% detection accuracy, high precision and recall rates, even when environmental conditions change. Once violations are detected, an alert buzzer goes off to ward off violators on the spot and further notify the relevant authorities. Vehicles parked in restricted areas for more than a minute activate the system, which captures a screenshot in high resolution (1920 x 1080) and keeps it in a timestamp-folder for later retrieval. The system processes more than 864,000 frames on a daily basis and requires about 80 Megabytes storage for daily evidence. These findings illustrate the scalability and utility of cost while providing a reliable solution to automated parking enforcement. Future enhancements will include license plate recognition with cloud data sharing for much richer functionalities.