Research On Character Recognition Technology In Vehicle License Plate Recognition System

The License Plate Recognition (LPR), which involves numerous subject fields, such as pattern recognition, artificial intelligence, computer vision and digital image processing etc., is one of a key technology of Intelligent Traffic System (ITS). The LPR system mainly includes three critical parts: license location, character segmentation and character recognition. This paper elaborates on the technology of image preprocessing in LPR system and the above three critical parts, and it also studies the character recognition technology at some depth. In this paper, according to the character features of Chinese characters and English characters, the author proposed two new classifiers respectively. The main works of the paper are sum-up as follows:At first, in the image preprocessing section, some common technologies and methods, such as the gray processing of image, the image enhancement, binarization, the edge detection, etc., are introduced. And secondly, in the license location part, the author analyzes the Chinese license plate (whose size is 440mm×140mm), and elaborates the characteristics of this kind of the license plate. The features contain the structure of the plate, the distribution of characters, the texture, the color information and the spectral information, and so on. These features can be used to locate the license plate. And some common location algorithms, such as the texture characteristics analysis, morphological methods, neural network approach and the wavelet transform etc., are introduced in this section. In the third part, the author analyzes the architectural feature that can be used in character segmentation, and recommends three segmentation algorithms of character---the method based on kerning features, the method based on connected domain features and the method based on the recognition results. At last, the author gives an overview of the character recognition technology in summary. For the common technology of character recognition, such as template matching algorithm and the BP neural network method, the author also makes some introduction. Especially for the BP neural network method, which plays an important role in character recognition, the author recommends it at length. In this section, on the basis of analyzing the code rule of Chinese license, the Chinese character recognition method based on connected and the English character recognition method based on the concavity of the image are put forward.