Detection of traffic rule violation in University campus using deep learning model

Implementing and monitoring traffic rules are essential for reducing accident rates and traffic rule violations. The automated traffic rule monitoring system ensures strict adherence to the traffic rules with low human effort. The system proposed identifies and acknowledges the two-wheelers violating the traffic rule regarding triple riders. The intersection points in the University Campus act as the data collection centres and collects the data through live recordings captured by surveillance cameras. The proposed system consists of detection and automatic number plate recognition. To implement this system, the Darknet framework is used, which is based on You Only Look Once (YOLO v8) for identifying two-wheelers with triple riders. The Depth estimation algorithm is used to detect vehicles and persons, which can accurately detect near and far objects. The vehicles are classified as Violator or No Violator. The Connectionist Temporal classification algorithm is used to classify the vehicles as violators or no violators. The two-wheeler classified as a Violator is stored in the database along with the vehicle license plate number, which can be penalised by traffic monitoring authorities. The implementation results show that the system is viable, efficient and reliable. Thus, make the two-wheeler follow the traffic rules properly, reducing the chance of irresponsible driving. The self-generated dataset detects the traffic rule violation and license plate number extraction. The system is trained on the considered dataset, and the best weights are used to implement the model. The proposed model achieves 94%, 96% and 97% performance for detecting triple rider, license plate and motorcycle detection.