The Research Of License Plate Character Recognition System

The paper introduces a part of license plate recognition system, and the aim is to set up a license plate character recognition system through using software completely. Import of the system is gray license plate image, and output of the system is a string of character that can be editted. This system includes license plate image binarization, license plate image pretreatment, segmenting character and the recognizing character.Segmenting character and the recognizing character deal with binarization image, so binarization algorithms decides veracity of character segmentation and recognition. The paper makes use of bimodal histogram and puts forward improved binarization algorithms . The algorithms is accord with the maximum error probability Bayes decision rule.In the course of pretreatment, the aim of smoothing gets two angle parameters that is obtained by median filter, verge detection and Hough transformation. We may decide if license plate has been distorted, according to two angle parameters. Correction algorithms is introduced in the paper and is a composite transformations matrix. There are four processes in correction of license plate image, consisting of parallel motion two times, rigid body rotation and non-rigid body rotation one time.First character segmentation adopts vertical projection diagram technique, after pretreatment of character segmentation has been done. Second character segmentation is based on result of first segmentation, according to knowledge about license plate. Many things taking place in the course of segmentation are discussed in the paper. After character image that has segmented successfully is standardized, we will get engenvector from it by using gridding method. The paper introduces how finds structure characteristic through making use of four operator. Finally,template library is composed of much engenvector. Further check in confusing character is realized through making use of contour engenvector.It is proved by experiment that the license plate character recognition systemhas good recognition rate and practical value.