Character Recognition Research And Application

With the rapid development of Chinese economy, the Intelligent Transportation System is fitted for the high requirements of road traffic management. License Plate Recognition System as an important part of ITS, is researched at home and broad.License Plate Recognition System is divided into our parts:the license plate location, character segmentation, feature extraction and pattern recognition. Our focus is the latter two parts.This paper presents a character feature extraction method giving stable characteristics of the characters and restraining disturbance mostly, such as brightness changing, rotation and deformation of character and strokes adhesion. The method process is, firstly getting edge information and gradient angle information for gray images, secondly forming part descriptions of strokes from sub image divided by sub-block method, finally outputting the vector of stroke feature.This paper gives an image pre-processing methods composed of Second Linear Interpolation Algorithm, Canny Algorithm and Sobel Operator, to meet the requirements of feature extraction.The effectiveness of character extraction method shows on right recognition partly, also the product of method-feature vectors directly. Euclidean distance is often used as the tools to evaluate effectiveness of method, but it is powerless in high dimensions. So this paper gives a similarity including complete difference concept, and a novel similarity function based on City-Block Distance, its properties, differences and advantages with Euclidean distance and other similarity functions. This paper sets up a series of evaluation, such as the average similarity of the same character, the average similarity of the different characters, the maximum similarity of the different characters, and the average divided marginLicense plate recognition system is designed and realized. The correct rate of system is more than95%in practical applications.