基于边缘颜色对的车牌定位新方法

车牌定位是车牌自动识别系统中的一个关键问题．该文提出了一种新的基于边缘颜色对的车牌定位方法．首先进行彩色边缘检测，然后以每一边缘点为中心，垂直于边缘方向取一线形窗口，在窗口内检测边缘点两侧像素的颜色是否分别匹配车牌的底色与字符颜色，若是，则保留为候选车牌边缘点；然后进行形态滤波，剥离不符合车牌结构特征的区域，最后对候选车牌区域进行纹理特征的分析以确定真实车牌区域．该方法抓住了车牌背景与字符具有固定颜色搭配的重要特点，综合利用了车牌的结构特征和纹理特征，提高了车牌定位的可靠性．对各种条件下拍摄的163幅含有车牌的图像应用该算法，定位准确率达到98．2％。

A Novel Approach for Vehicle License Plate Location Based on Edge-Color Pair

Locating the vehicle license plate plays an important role in the vehicle license plate automatic recognition system. A novel locating approach based on the edge color pair is presented in this paper. Firstly, the edges are detected in a color car image; secondly, a line shape window is made for every edge pixel, and its direction is perpendicular to the direction of the edge and its center is located on the edge pixel, and then the color pattern of the pixels on both sides of the edge point in the window is investigated and the centric edge point of the window is reserved when the color pattern matches the combination of the background color and text color of the plates. Whereafter, a morphological filter is applied to the candidate binary image for removing the regions without the structure feature of the license plate. Finally, the license plate is extracted correctly from the candidate regions by analyzing the texture feature of the plate. The proposed method focuses on matching background color and character color in a license plate and combines its structure feature and texture feature. The experiments on 163 car images that were taken under various conditions show the extraction rate of 98.2%. Integrating color edge detection, edge color pair, mathematical morphology and neural network into our method,the approach offers robustness when dealing with noisy car images, car images in variant lighting conditions and car images with skew number plate.