Traffic Management System using Object Detection

Ensuring road safety is a major responsibility, but manually spotting red light violations can be tedious andprone to mistakes. That’s where our Real-Time Traffic Management System comes in, using advancedcomputer vision and deep learning to automate the process. At its heart, the system relies on YOLOv5 (YouOnly Look Once, Version 5), a powerful object detection model that quickly identifies vehicles fromrecorded footage. Once a vehicle is detected, SORT (Simple Online and Realtime Tracker) follows itsmovement across multiple frames, ensuring accurate tracking of its path. To catch red light violations, thesystem uses template matching, allowing it to recognize when a traffic light is red and detect if a vehicle hascrossed the stop line. When a violation happens, the system highlights the vehicle with a bounding box,tracks its movement, and sends an alert to enforcement authorities. This automation not only boostsefficiency and accuracy but also reduces the need for human intervention, cutting down workload andpotential biases in manual monitoring. Designed for practical use, the system seamlessly processes prerecorded footage, making it highly adaptable and scalable for modern traffic management. We also compare YOLOv5 and YOLOv8, analyzing how they perform in image-based detection. YOLOv8, the latest version, delivers better detection accuracy, faster processing, and improved feature extraction. By testing bothmodels on the same dataset, we assess which one is more effective at detecting red light violations andenhancing traffic law enforcement.