Design And Realization Of Vehicle Cognition System

Recent years, as the increase of vehicles, the vehicle management becomes more difficult. It needs Intelligent and systematic management. Vehicle recognition system is a important issue of Intelligent Transportation Systems. It has spacious application and development prospect. Until now, researchers concentrated on the rude classification of vehicles. In this work, I not only have done some work on the rude classification, but also have done some research on the subdivision of the vehicle.There are three difficulties in vehicle recognition by using digital image processing methods. First, it is difficult to extract outline of vehicle in a complex background. Second, the outline of vehicle is changing when it is in different area or different angle of view. Third, sometimes, the outline of vehicle is not enough to classifying vehicle type.In this work, the colors parameters are fully used in background subtract. In the result image the contrast is enhanced and vehicle area is distinct. In order to fill the gap and wholes of the vehicle area, some consequent images are stacked in the same size. The perfect outline of the vehicle is extracted from the stacking image. There are 7 Invariant Moments, which is constant in image movement, rotate, mirror and zooming. The 7 Invariant Moments are used to classify the vehicle type and the recognition is accuracy.Although it is successful in the rude vehicle classification by 7 Invariant Moments, we have to use more useful vehicle information in the subdivision of the vehicle. We can divide the vehicle image from the resource image by the outline of the vehicle, and then image edge can be detected with boundary operator. In view of the 2D-Gabor filter can cover both microcosmic and macroscopic feature,2D-Gabor filter is used to describe the image edge. By using 2D-Gabor filter we can get a higher recognition rate than using the 7 Invariant Moments.In addition, I realized a simple vehicle recognition system in my computer. The software can capture video and play back by using DirectShow. The area within a short distance of camera Optic Axis is set as the effective coverage. When the vehicle is in the effective coverage, it's image is considered as the integral and it's shape is remain unchanged.