基于颜色特征的车牌快速定位

Fast License Plate Localization Based on Color Features

提出了一种基于颜色特征的车牌快速定位算法,该算法充分利用车牌颜色相对固定的特点,首先根据原始图像得到一组特定的色彩距离图谱,通过自适应熵阈值的选取快速分割出车牌的候选区域,然后再根据车牌的纹理特征对候选区域进行筛选以得到车牌的精确位置。针对不同的背景和不同种类的车辆,抽取了3106幅图片进行测试,其中有69幅图片未定位出车牌,定位成功率为97.8%,平均定位耗时为29ms。

This paper presents a fast vehicle license plate location method based on color features,which makes full use of the characteristic that license plates have relatively steadfast color.First we get a group of color distance maps for given colors,and the candidates of license plates can be fast and roughly segmented by choosing an adaptive entropic threshold.Then we validate the candidates according to the texture feature of license plates.In this way,vehicle license plates can be accurately located.In the experiment,3 106 images taken from various vehicles and different scenes were employed,of which,69 images were failed to locate the license plates,the license plate location rate of success is 97.8%,and the average time that each location needs is 29ms.