The Vehicle License Plate Automatic Recognition System

This thesis study vehicle plate the License Plate Recognition concerning the characteristics of Chinese. The License Plate Recognition system is composed of The License Plate Location, Character segmentation and Character Recognition.The License Plate Location: a method of color License Plate Location which based on YCbCr color space and projection method, this method extractive the Cb component of YCbCr color space. Dilate and erode is used for this component image which is a gray image. The License plate image created by Projection method used for the processed image.Character segmentation: a method round scan algorithm based on projection method for Character segmentation. In this part, license plate binarization method under nonuniform illumination is put forward. this problem is always met in license plate recognition system which must work under complex illumination, in the circumstances, It's difficult to binarize the license plate image. A new method is presented, Experimental results show that this algorithm is robust in dealing with nonuniform illuminated. At last this paper applied round scan algorithm to processed image. This method is propitious to two part and three part character.Character Recognition: In this part, the PCA (Principal Component Application) method is applied to extract eigenvalue and eigenvector of image. MDC (Minimum Distance Classifier) is applied for character recognition. This paper divided character image of image sample library in 5*5 segment to reduce dimension, PCA is applied to these segments in order to get 25 dimension eigenvalue and eigenvector. At last Appling MDC for input image achieve character recognition.In video surveillance, the car plates of interest are often of small size. Image restoration is important factor affecting car plate recognition. In this paper, we introduce a restored arithmetic of low-resolution car plate image based on eigentransformation. This method views restoration as eigentransformation between different image styles. This method uses Principal Component Analysis (PCA) to fit the input low-resolution image as a liner combination of the low-resolution car plate images in the training set. The high-resolution image is rendered by replacing the low-resolution training images with high-resolution ones, while retaining the same combination coefficients. Experiments show that the restored car plate images are not only very helpful for recognition by human, but also make the automatic recognition procedure easier.