The Research Of License Plate Identification Algorithm

License plate identification has become an important part of the "peace city" construction, road traffic management for help, they are not only applied to a red light shot, speeding violation paparazzi, vehicle travel time statistics, highway traffic management application of road traffic application, but also applied to public security safety monitoring such as traffic security monitoring license plate identification, vehicle wanted criminal case detection, auxiliary have been widely used.The key to LPR technology mainly consists of three parts, the most important parts include license plate localization, character segment and character recognition, and the license plate localization of the image processing and license plate segmentation of the tilt correction before related introduction. In this paper the three key technologies are further exploration and researched, and put forward the effective method, the concrete includes the following contents:According to the plate region hard to locate, combined with prior knowledge of vehicle license plate, and put forward a kind of regional localization algorithm. The projection method, based on the license plate structure characteristics, to locating the license plate boundary, then locating the license plate, the results show that this method has less time consuming, the characteristics of high accuracy.Based on the analysis of the global threshold value method and local threshold value method on the basis of their respective advantages and disadvantages, this paper proposes two party with the combination of binary image algorithm and the simulation results show that this algorithm is effective to solve the fuzzy character, adhesion brings the segmentation problem, still can effectively remove interference, and has good robustness, segmentation speed, high accuracy.Based on the analysis of the traditional template matching character of recognition method’s deficiency,put forward another kind of improved algorithm. The improvement of the algorithm is based on the BP neural network’s design, make full use of the character background region for matching the effective to solve the problem, and gives the solution of similar character recognition. According to the complex type of license plate character, stroke rich characteristic, this paper puts forward to a single character recognition, and the experiment results show that the method character average recognition rate is93%.