Research And Implement Of Vechicle’s Multiple Features Recognition Algorithm

Vehicle identification is an important part of intelligent transportation systems regions. With the improvement of people’s living standard, the number of vehicle has increased dramatically, which lead to city’s traffic system’s pressure is becoming more and more big. The system can be used for extracting vehicle information automatically, efficiently and intelligently, which plays an important role in quickly solving problems such as traffic accidents, violations and congestion. In this thesis, we will mainly studies two aspects of the vehicle identification: license plate recognition which mainly improve the location algorithm of license plate and license plate tilt correction algorithm, vehicle-logo recognition which is a complete recognition system from location to recognize.(1)This thesis studies a license plate localization algorithm by fast feature pyramids. It applies Fast Feature Pyramids algorithm combined with machine learning AdaBoost classification algorithm to locate the license plate,aiming at the low accuracy of license plate localization and higher false detection rate of car license plate in complex and extreme environments. The algorithm improves the accuracy of license plate location efficiently, and reduces the false detection rate of car license plate, which has great robustness.(2)This thesis studies a license plate image tilt correction algorithm which based on the TILT algorithm. According to the principle that the symmetrical texture information are characteristic of low rank, we propse a new license plate tilt correction algorithm which solve the difficultes in dealing with large tilt angle and shear correction of raditional algorithm. The algorithm is effective to expand the scope of license plate tilt correction angle, and a good effect for shear correction.(3)This thesis proposes a secondary logo location algorithm which based on visual signature. According to the front part of the logo and the license plate position relationship, the algorithm realize the logo coarse localization algorithm which based on license plate location algorithm,then using visual signature algorithm to accurate positioning. In complex scenarios the algorithm effectively realize the vehicle-logo precise positioning,which having higher accuracy and lower error detection rate.(4)This thesis studies a vehicle-logo recognition algorithm which based on histogram of oreiented gradient features and projection features combination of support vector machines classifier. Based on logo position algorithm and the whole vehicle-logo recognition system for real-time requirements, we choose a suitable vehicle-logo recognition algorithm of this thesis. The algorithm not only ensure identification accuracy but also effectively achieve the vehicle-logo recognition system’s real-time demand.