Research Of Vehicle Identification Technology On Many-core Platform

With the rapid development of the transportation, ITS(Intelligent Transport System) has obtained extensive concern. As the key technology of the ITS, vehicle type recognition technology and vehicle color recognition technology have become a kind of research focus. In recent years, vehicle type recognition technology is developed rapidly, which is not only widely used in the field of ITS, but also springs up in the field of image processing, pattern recognition, computer vision and so on. Automotive body color recognition technology fills the gap in the information of vehicle color feature in the intelligent vehicle recognition system, complements the features information of the vehicle recognition technology and helps the police attack criminal behaviors on car license plate.This paper is mainly focused on the researches of vehicle type recognition and vehicle color recognition. Firstly, the author studies the location algorithm of the license plate, and locates the space position of the license plate by using location algorithm of license plate based on HSV color space. Then according to the positional relation between the face and license plate in space, she locates the vehicle face area and matches the corresponding vehicle type by using the experiential and statistical methods. This paper introduces image matching methods based on the SIFT algorithm. It proposes vehicle type matching methods with two-way matching to improve the accuracy of vehicle type matching. The author uses a parallel design for SIFT algorithm. Finally, the paper achieves SIFT algorithm parallelization on the CUDA platform, meeting the real-time requirements of vehicle type matching. In the study of vehicle color recognition technology, the author gives full consideration on the positional relationship in space between the vehicle hood and the license plate to locate the vehicle color recognition area. She proposes vehicle color recognition method based on the color histogram in the HSV color space, improving the accuracy of vehicle color recognition.