License Plate Location System Based On Quotient Space

Research on vehicles' plate license recognition system (LPRS) involves techniques such as digital image processing, pattern recognition and artificial intelligence etc. License plate recognition which is a difficult problem is solved by the synthetic application with the method and the thought of each domain that be listed above. The research and application of each domain is indispensable to the research and improvement of license plate recognition. LPRS consists of two modules in general, which is license plate locating and character recognition. License plate location is an important part of LPRS. The efficiency of license location method is a bottleneck in system performance, therefore this article aims at the car license location question, which makes the analysis and the improvement, and realizes entire process.The primary work of the thesis is listed below:1, Firstly, this thesis introduces the basic theory and method of image segmentation, as well as their characteristics and practical application. Some common methods and characteristic about license plate localization are introduced. Secondly, it introduces the quotient space granular computation theory, quotient space granular computation theory gives us a theoretical basis for understanding, describing, analyzing, researching and solving complex problems.2, In the pre-processing phrase, we improve the quality of the input image by adopt segment linear function to stretch grayscale images on the basis of the original image analysis and testing. Then the fuzzy edge detection algorithm is used, which has better performance to detect image edge. The extracted edge information of the image is used as license plate location algorithm input information.3, The license plate location question is described by quotient space granular computation theory, and a license plate localization algorithm is designed by hierarchical thought of quotient space granular computation theory. The result of the experiment shows that the method increases the accuracy and adaptability of locate. 4, In the binarization phrase, in regard to the affect of illuminance, we used a algorithm which can improve robustness of binarization method against illumination to refine the binarization result. In plates skew correction phase, in regard to some traditional algorithm makes it difficult to achieve good results when the license plate characters region have no border or the arrangement of characters specially. we used a novel license plate slant correction method based on principal component analysis to achieve a better correction results.