Space invariant vehicle recognition for toll plaza monitoring and auditing system

Space invariant object recognition is one of the difficult problems of pattern recognition and has many potential applications. In this paper, a vehicle recognition system is proposed for toll plaza monitoring and auditing. This system recognizes the type of approaching vehicle such as truck, bus, car etc. irrespective of geometrical distortion of vehicles such as scale and rotation. Maximum Average Correlation Height (MACH) filter is used to obtain out-of-plane rotation invariance and Log r-o Mapping is employed to achieve in-plane and scale invariance. Five classes of vehicles are defined: trucks, buses, wagons, cars and motorcycles. MACH filter is trained on log r-o maps of each class of vehicles for range of orientation and correlated sequentially with the log r-o map of region of interest. The type of vehicle is recognized from correlation results. The system analyzes the real time toll plaza video and generates audit report indicating the total count of each type of vehicles during a specified time and the resulted toll collection according to their respective charges. The system also generates monitoring alerts as per given decision rules.