Vehicle Abnormality Detection and Classification using Model based Tracking

In this paper, we present a novel approach for detection and classification of abnormal vehicles in urban highways. We considered four types of abnormal vehicles such as near pass, illegal lane crossing, slow moving and long time stopped vehicles and which are detected and classified based on vehicle tracking and speed information. The model based tracking is used to detect, near pass and illegal lane crossing vehicles. The slow moving and long time stopped vehicles are detected based on speed information computed from model based tracking information. Since, the detection accuracy is mainly depend upon reliable and most efficient tracking, we adopted model based tracking which able to track the vehicles efficiently under various situations such as shape and appearance variations in road crossing, illumination variation and complex background. The two video sequences selected from iLids and GRAMRTM are used for experimentation and results are evaluated based on precision, recall and fmeasure. The experimental results demonstrate that our approach achieves highest accuracy for detection and classification of abnormal vehicles in urban highways.