LAUSR

Traffic violation monitoring and control is a major concern in India due to excess crowd, increasing commuters, bad traffic signal management, and rider mentality. It is obvious that physical traffic police-based monitoring alone is insufficient to monitor such large traffic volumes and simultaneously track violations. This has led to many violators going unnoticed. The violators, in turn, cause more serious mishaps on the road resulting in danger to their own life as well as to other’s life. Thus, there is a need for incorporating Artificial Intelligence (AI)-based techniques to eliminate manual intervention for the detection and catching of violators. In this paper, we propose a system to automatically detect two-wheeler violations like not wearing a helmet, usage of a phone while riding, triple riding, wheeling, and illegal parking for Indian road scenarios and eventually automating the ticketing process by capturing the violations and corresponding vehicle number in a database. We propose using a custom trained Yolo-v4 + DeepSORT for violation detection and tracking and Yolo-v4 + Tesseract for number plate detection and extraction. This implementation obtained a mean average precision (mAP) of 98.09% for violation detection and an accuracy of 99.41% for number plate detection on the test data. Further, the system detected 77 out of 93 violations with zero false positives in real-life scenarios. Thus,showing that the traffic violation system developed can be used to automate traffic violation ticketing. The developed system would be particularly useful in deriving various safety-related policies and will help to enforce strong regulation of traffic rules and build towards a smart city ecosystem via the automated AI-based traffic violation and ticketing system.