License Plate Location Based On The Constructive Learning And Recognition System

With the increase of the possession of car and the requirement of transportation, the problem of traffic has been more serious than before. Confronted with such situation, people begin to research how to use ITS (Intelligent Transportation Systems) which based on the technique of computer to administrate and control the transportation effectively. In an ITS, the location and recognition of vehicle license plate plays a very important role and can be used in expansive application ranges and has a promising future.In this thesis, we combine the theory of Neural Network and traditional image processing technique, analyze and research the topic of the location and recognition of vehicle license plate based on structural alternative covering algorithm, and try to locate the car plate even there exist lots of distributive factors in the image and recognize the characters finally. In total, the main work and creativity in this treatise can be summed up four points:1) After analyzing the traditional means to recognize the car plate with texture, we propound the conception of dubious region and extract the dubious regions from original image to process in later steps. Thus, the location of car plate is decomposed into a procedure from rough to accurate;2) We use structural alternative covering algorithm and abstract the respective features of each dubious region to study and construct a neural network in the location of car plate. Via this neural network, we can eliminate those regions which contain no plate and then use color information to modify the correct region and find the accurate position of car plate finally;3) In the progress of recognition, we apply the structural alternative covering algorithm and only use some basic techniques to process the image. In other words, we depend on the strong ability of classification of NN to recognize the characters of a car plate;4) Under the precondition of inadequate samples, we introduce the technology of additional sample in the progress of characters recognition which uses experiential knowledge to construct some sample artificially and then inserted these samples into the sample set in order to improve the performance of network.The result of the experiments proves that the method which we use in this paper can locate and recognize characters of car plate effectively with a high recognition rate and has a practical usage.