lab颜色空间和形态学处理相结合的双行车牌定位方法

车牌定位是车牌识别系统的前提和关键。针对双行车牌均是黄色车牌,且比较脏的特点,首先将图像由RGB颜色空间转换成Lab颜色空间;再利用Lab颜色空间的通道a提取出图像中的红色和绿色区域,通道b提取图像中带干扰的黄色区域,然后将二者相减,提取出图像中的黄色区域,并可同时去除背景及车身的大部分干扰;最后利用形态学处理滤除噪声等影响,粗略定位出车牌候选区域,再结合图像的纹理特征如面积、长宽比和连通域内像素个数最终定位出车牌区域。该方法对复杂环境下的双行车牌能实现快速准确定位,受光线、背景环境影响较少,同时对脏牌、污牌也能达到准确定位的目的。

Two-line license plate detection method based on CIELab space and morphological processing

License plate detection (LPD) is the premise and key of license plate recognition system. For two-line license plates are yellow and dirty, this paper will first transform the image from the RGB color space to CIELab color space; Then use b channel of CIELab space to extract yellow areas with interferences, use a channel to extract red and green areas. And then subtract the two extract images to get the yellow areas of the image, at the same time can remove most of the interferences of the body and background; Finally utilize morphological processing to filter out noise which can get license plate candidate areas, use texture features such as area, aspect ratio and the number of pixels in connected domains to locate the license plate. Experiments show that this algorithm can fast and accurate locate the license plate in complicated environment, and have less affected by the light, background environment. Even dirty cars can also achieve the purpose of accurate position.