Integrated Traffic Violations Detection System for the Highways of Bangladesh

Abstract

Effective traffic violation detection on highways is crucial for upholding traffic law and order, guaranteeing a smooth flow of traffic, and minimizing congestion and delays. The goal of this thesis is to create an integrated system for simultaneously identifying various traffic violations on Bangladeshi highways, with an emphasis on vehicles that aren’t allowed to travel on them. To achieve greater accuracy in identifying offenders, the system combines various violation detection approaches and makes use of YOLOv8, the most recent version of YOLO. In this research, four traffic violations are detected, monitoring vehicles which are illegal on highways like CNG driven auto-rickshaws and easy-bikes, over-speeding, illegal road crossing on highways and wrong way driving. A new dataset containing images collected from the roads of Bangladesh were created for detecting illegal vehicles. For detecting the other three violations, YOLOv8 and its native tracker is used. The three violations viz wrong way driving, over-speeding and illegal road crossing detection systems were combined using a logical framework. The integrated traffic violations detection system performed with an accuracy of 95 percent which proves the efficiency of the system on Bangladeshi highways. This study provides an efficient integrated system for monitoring traffic violations on highways and serves a tool for the assistance of highway traffic authorities