Research On The Key Technique Of Vehicle-logo Identification

With the rapid development of science and technology and growing need of traffic cause, intelligent traffic system has become a hot topic. As an important part of intelligent traffic system, the vehicle identification system has been widely applied in the nonagency parks, automatic charging, traffic accident treatment, vehicles peccancy automatic recording and other fields. Therefore, the fulfillment of the vehicle identification system has enormous economic value and practical significance.As an important respect of vehicle identification, vehicle-logo identification technique is very practical. It takes the digital images and video signals as research objects, and learns the vehicle's brand information through image processing and automatic identification. Vehicle-logo identification system consists of two key techniques, which are vehicle-logo location and vehicle-logo identification. Most of the existing vehicle-logo location techniques are based on the license plate, so the license plate location technique is a key for the vehicle-logo location. In this thesis, the author will deeply study the technique s involved in license plate location and vehicle-logo identification.The vehicle-logos locate on the medial axis, which is an important feature of vehicle-logo. Due to the positional relationship between the vehicle-logo and license plate, the license plate location technique plays a key role for the vehicle-logo location. Because of the complexity of image background, and the inscrutability of license plate position and image quality, it is difficult to locate the license plate during the process of license plate image acquisition. This thesis puts forward a new method of license plate location which bases on syndromes. This method analyzes the license plate area as a whole instead of several characters. It coarsely locates the license plate through the structure features, then rejects the false license plate area which is not similar with the texture of license plate, and finally accurately locates the license plate based on the feature of edge-color pair in the license plate area. The experiments show that, The method of this thesis has higher accuracy and stronger robustness.Then apply the PCA technique for data reduction, and calculate the invariant moments as feature vectors, the last application of BP neural network technology to identify the vehicle-logo. The experiments show that the proposed method of identification vehicle in this thesis has rapid identification speed.