Road safety is put at risk by violations of traffic rules including riding a motorcycle without a helmet and using cycle lanes improperly. A deep learning-based framework for the automated real-time identification of these violations is presented in this research. The suggested system uses advanced object detection and tracking algorithms in conjunction with spatial reasoning to detect bicycles riding outside of approved cycle lanes and motorcyclists without helmets. To improve detection accuracy, the system uses bounding box modifications, centroid-based relationship, and region-specific filtering. Additional elements, such as speed and directional analysis, add context to the observed violations.In addition to providing visual feedback and keeping track of cumulative counts, the system performs excellently in identifying and reporting violations. The architecture is flexible and can be expanded to handle a wider range of traffic violations. It is made to function smoothly in a variety of urban traffic situations. The suggested technique reduces dependence on manual monitoring by automating violation identification, and thus helps traffic management authorities improve road safety. In order to verify and improve the system, future development will concentrate on increasing functionality, enhancing edge device efficiency, and carrying out realistic deployments.

Computer Vision Traffic Surveillance YOLO Vehicle Speed Detection Direction Detection Helmet Detection Lane Violation Non-ANPR Cameras