The Research On Car License Plate Location And Recognition

License Plate Recognition (LPR) is one of the most critical research topics of Computer Vision, Image Processing and Pattern Recognition applying in Intelligence Transportation System (ITS). It is an important part to realize the intellectual traffic management, and it holds the balance in the project management of the freeway, city transportation and park management.Vehicle License Recognition System based on video images is a unitary intelligent system itself, which is composed of Image Collection, Plate Location, Slant Correction, Character Segment and Character Recognition. This paper analyzes the characteristics of the car images and the license area characteristics, utilizes the pre-processing methods in many points of view to different taches, and then, explores the reasonable solutions to realize Location and Recognition combining with the knowledge of Pattern Recognition and Computer Vision.The research work in this paper mainly contains three parts:(1) Plate Location. The paper synthesizes the license area information, as well as texture, color and geometry characteristics, realizes the location in two-steps. At first, do gray-level processing to extract edge-compression areas as the candidate plates. Then, to the small color images cut out, utilizes color space processing techniques based on HSI model. The operation reduces the computation quantity, and satisfies the request of expeditiousness. Moreover, HSI color model separates the connection between brightness and color; the extracted characteristics have the traits of veracity and stability. It can decrease the plate's wrong-location, and provides a solid foundation to Character Recognition.(2) Character Segment. The located vehicle license has always been some slantwise, this paper uses the methods on the basis of Hough Transform and Mathematics Morphology to do slant correction (it is not this paper's work). Character Segment is realized in the methods of pixel-value change and horizontal-vertical projection techniques combining with the inherence characteristics of the license area.(3) Character Recognition. To the recognition of Numbers and English Letters, this paper extracts the skeleton of the character by use of morphologic thinning, codes the skeleton according to the 8-directions chain code to describe the configuration of the character, puts forward the conception of Simple-code and constructs its characteristic, and then designs multilevel classifier corresponding with different rules. Chinese recognition is implemented in methods of template match, and its match characteristic is picked up by structure operators.The experimental results reveal that the arithmetic proposed in this paper has good stability. Especially in the aspects of Plate Location and Character Recognition, it gains a great improvement in running efficiency and accuracy compared with conventional methods.