The Traffic Light Violation Detection System is an innovative solution that relies on computer vision techniques to improve traffic management and road safety. This project seeks to develop an intelligent system capable of automatically detecting and monitoring traffic light violations at intersections. Using computer vision algorithms, the system processes video signals from surveillance cameras installed at traffic lights. The proposed approach includes several stages, such as video preprocessing, object detection, and violation classification. In the preprocessing stage, video frames are analyzed to extract relevant information. This step contributes to improving the overall accuracy of subsequent stages. The system uses advanced object detection algorithms, such as Ultralytics YOLOv7 and SORT (Simple Online Real-Time Tracking), to detect vehicles within video frames. By accurately identifying objects, the system can analyze their behavior and compliance with traffic rules. The classification stage determines whether a vehicle has violated a traffic light rule—for example, if it ran a red light—to achieve robust performance. The project uses a comprehensive dataset of labeled traffic scenarios to train and enhance deep learning models. We used the COCO (Common Objects in Context) dataset, which contains over 348,000 images. We also trained a custom model to detect red, yellow, and green lights, as well as pedestrian crossings.
