SMART TRAFFIC VIOLATION DETECTION SYSTEM

Authors:

Priyanka Balaga, Javvadhi Dhanush, Piradi Balu, Vincent John Mark Dondapati, Andhavarapu Dhanush

ABSTRACT:

Certain infractions are the main causes of accidents as well as traffic jams on public streets. Among these, the most common three are riding a motorcycle without a helmet and having three riders on one motorcycle. Manual monitoring of traffic violations is not feasible because of the tasks involved and potential errors made by humans. Thus, the objective of this project is to build an AI automated traffic violation detection system using Closed Circuit TV/video captured in real-time. The AI Detection System will be able to keep track of every violation that happens in any one area at all times, using computer vision technology to analyse the video being received from the cameras and Provide evidence of suspected violations in real-time to law enforcement with a report as appropriate. The model is based on YOLOv8 object detection and has been developed using the Roboflow Data Set. Additionally, the performance of the model will be improved using several types of augmentation, depending on the conditions in which the video is captured (changing brightness, zooming, cropping, etc.). Google Colab with GPU capability has been used to improve the speed and accuracy of the YOLOv8 training process. The new AI based detection system will be able to integrate with existing security camera systems and can be used as part of an intelligent city solution with improved vehicle/traffic monitoring capabilities.