Research Of The Vehicle License Plate Automatic Recognition

This paper discusses on the topic of my researches in Vehicle-License-Plate automatic recognizing technologies, it will mainly about the applications of Image Processing technologies and Pattern Recognition technologies which applied in LPR. A typical LPR system will include image collecting part, plate location and segmentation part, character segmentation part, character recognition part and result processing part etc. This paper is mainly about three parts that of : the first part that Vehicle-License-Plate automatic location and segmentation, the second part that License-Plate-Image processing and character images segmentation, and the last part that License -Characters recognition.In the first part, we present a kind of segmentation method of Vehicle-License-Plate images based on color clustering and mathematical morphology. Based on ocular uniformity,we process our color clustering in Munsell color space using the NBS distance. Then,with different structure elements,we take a series of operations to the image. At last, we find the location of the plate through the two-dimension histogram of the image.In the second part, we do some processings to the License-Plate-Image, include to filter the high frequency noises through a method based on wavelet transform, we weaken the wavelet coefficients in different bands incoordinately; we make the image binary by thresholding through a combine of a iterative method and a OTSU method; we adjust the size of the image to a normal value and we correct the skew of the image by detect out a angle through Hough transform. At last, we segment the character images through two-dimension histogram and use the experience of the ratio of the width and the height of a Vehicle-License-Character.In the three part, We recognize the characters based on the features picked out by K-L transform. Combining a method to make the error probability's high limit minimum, we upgrade the method to a high accuracy.The experimental result shows that the proposed approach is gratifying. We got a high recognizing rate and reliability in the recognition of license characters.