Research For Recognition Of Moving Vehicle Type

Recognition system of vehicle mainly consists of two key technologies: recognition of vehicle_license_plate (VLP) and recognition of vehicle type. It not only finds wide application in ITS, but also is a hot point of research in computer vision, image processing and pattern recognition. So its related technology is attended prevalently. In our paper, we pay attention to the location and recognition of vehicle logo. The methods to deal with above problems are proposed in the paper and testified in the experiment. The paper includes five parts: 1. Research of object detection and recognition algorithm: object detection and recognition algorithm are key problems in object recognition, and recognition of moving vehicle type is a typical system of automatic target recognition. In this section, generalization and analysis of the image segmentation and feature extraction and object recognition are emphasized to be studied. 2. Vehicle detection and recognition of vehicle shape and color: We also explore other important technical problems in recognition system of moving vehicle type, such as the detection of moving vehicle based on video images, recognition of vehicle type according to shape, size and color. Because the background of vehicle video detection is absolutely static or unchanged, we use difference method to detect vehicle. Based on the comparative research on the color difference of different color spaces, we find out some practical color space and color difference formulas. Further on, we improved the recognition precision and its robust by means of color normalization. 3. VLP location algorithm: VLP location in a vehicle image mainly assists the succeeding steps for vehicle logo location , which will directly affect further process and is a key step to improve the precision of vehicle logo location recognition. Its difficulty is that because of noise, the segment threshold of norm method is hard to adjust, resulting in errors in VLP location. So, an adaptive energy filter and a edge project method are proposed in the paper. Experimental results show that both are a fast, robust and accurate method of VLP location and meet needs of real-time system. 4. Vehicle logo location algorithm: Vehicle logo location is an important part of research system for recognition of moving vehicle type. While the texture feature, shape and size of every vehicle logo are different, It can consume much time if we adopt the conventional object location algorithm, such as relation match. A fast an robust vehicle logo location algorithm is proposed in this paper. The approximate scope of vehicle logo is first confirmed by preexistent knowledge, then its finer position is found by energy calculation and morphology operators. According to this process, we propose a practical method of vehicle logo location from coarse to fine. Experimental results approve the fast and efficiency of the method. 5. Vehicle logo recognition algorithm: After location, vehicle logo recognition becomes a 2-D shape recognition problem. But being subject to illumination and noise, it is hard to get accurate results by common methods. The range of second location is larger than that of vehicle logo to improve the location stability. To improve the accurate of recognition, a vehicle logo recognition method which integrates template matching with feature matching is proposed. Based on the first recognition of template match which calculate relative index using vehicle logo standard template, we get the accurate position of vehicle logo and use matchingvalue to recognize vehicle logo. The global feature of vehicle logo is extracted by edge direction histogram to match feature and recognize vehicle logo. Experimental results approve the accurate of the method.