Vehicle Classification and Violation Detection on Traffic Light Area using BLOB and Mean-Shift Tracking Method

Abstract

Obeying traffic regulations is an obligation for every driver. Although obeying traffic regulations is mandatory, there are many violations that occur on the road such as breaking through road markings or traffic lights. The police have implemented many innovations to reduce the occurrence of traffic violations, but they are still inefficient in handling traffic violations. In this research, a vision system was built to recognize vehicles (cars and motorcycles) and to recognize violations committed by drivers at road intersections. In this study, a vision system was built to recognize vehicles (cars and motorcycles) and to recognize violations committed by drivers at road intersections. Vehicle violation data is taken from CCTV cameras installed at road junctions. the system works for vehicle detection and traffic violation detection. The process of vehicle detection by searching for vehicle contours uses the BLOB method. Furthermore, in tracking vehicles using the Mean Shift Algorithm. To recognize road marking violations. We create a border as a reference in the CCTV video using image processing. When the traffic light sign stops, and the vehicle does not stop at the reference line, the vehicle is considered a traffic violation. When the vehicle has to turn, or go straight, it must be in the same direction as the reference line, otherwise it is considered a traffic violation. The detection results for motorcycles are 79% and cars are 71%. Meanwhile, the average result of violations on road markings is 58%.