Research On License Plate Character Segmentation

License plate recognition system is through computer vision, image processing and pattern recognition method to extract the characters information of image from the vehicle license plate, to determine the identity of the license plate, that is one of the important part to achieve intelligent traffic management. License plate recognition technology has important practical value in this fileds.License plate recognition system is an important research topic in the Intelligent Transportation. It includes three key components: license plate location, license plate character segmentation, and license plate recognition.In this paper, character segmentation technology is studied on the basis of consulting a large number of literature, we propose two license plate character segmentation methods.This paper proposed based on discrete cosine transform and identify the license plate character segmentation algorithm for feedback. First of all, image preprocessing was applied to the license plate; Secondly, vertical projection was used for rough segmentation of the plate characters; Finally, two-level feedback was employed for fine segmentation of the plate characters. The first level feedback was the feedback based on the character count and single character width, and the second level feedback was the feedback based on the character recognition. The experimental results show that the image preprocessing technology solves the license plate image with uniform-illuminated , contrast–poor ,declining; taken from coarse to fine segmentation method can correctly separate dirt, fade, character-touching and character-broken number plates and other serious effects of plate character image. Based on the feedback of characters segmentation in particularly, from the vertical projection for discrete cosine transform, application of its low coefficient of vertical projection to reconstructe image, eliminate the noise. On the basis of texture features better reflect the character license plates texture, improves the accuracy of character segmentation.Based on wavelet transform and identify the license plate character segmentation algorithm for feedback is proposed. The basic idea is still used by the coarse to fine, two-level feedback was employed for segmentation of the plate characters. Feedback in the second stage, wavelet transform is used to the vertical projection of binary images. Because hight frequency components determining the detail part, low frequency components determining the overall shape, so the low frequency components of wavelet transform are combined the vertical projection statistics with texture feature vector. Then feedback based on the character recognition are performed. The experimental results show that the proposed algorithm is robust to character segmentation when dealing with uniform-illuminated, contrast–poor ,declining , dirt , character-touching and character-broken number plates.