An efficient traffic management system to detect lane rule violation using Real-time Object Detection

Abstract

There has been an upsurge in the number of issues with Bangladesh’s present traffic control system. Hence, several accidents have occurred frequently. The two primary causes of a rise in the number of injuries are violations of traffic laws, such as illegal lane changes and excessive speeding. Here we have presented extensive research with an intention to resolve the current traffic management system using real-time object detection. In our proposed system, an edge node will detect the lane-based rule violation and send the data to the nearest intermediary node. Afterward, License plates as objects will be detected using YOLO object detection executed in the intermediary computing device. Finally, extracted license plate images from the intermediary nodes will be sent to BRTA traffic servers to detect the violator’s Bangla license plate number using pytesseract. We have built a data set of 1450 images for object detection and achieved an accuracy of 91%. Our system will assist the traffic control department in identifying those responsible for traffic rule violations and ensuring that the laws are strictly enforced.