Abstract: This paper presents a method of license plate location based on edge features filter and multi-decision mechanism under complex backgrounds. The edges of license plate are retained by using the edge features of target region. The candidate regions are acquired by the morphological processing. This paper adopts the multi-decision mechanism which combines the texture features and the color features of license plate to remove the pseudo plate regions. Experimental results of many images under complex backgrounds show that this method is effective and fast, which overcomes the sensitivity to image environment, illumination, color, and shooting angle.

Key words: license plate location, edge features filter, mathematical morphology, texture features, color features, multi-decision mechanism

摘要： 针对复杂背景下的车牌定位，利用目标区域的边缘梯度特性筛选出车牌边缘；选取合适的结构元素做数学形态学填充，得到车牌候选区域；融合车牌灰度纹理特征与颜色特征通过多判定机制剔除伪车牌区域，实现车牌的准确定位；通过实验对大量实拍的复杂背景下的车辆图像进行测试。结果表明，该方法准确率高、速度快，克服了传统算法对拍摄环境、光照条件、颜色、角度等较为敏感的问题，具有良好的定位效果。

关键词: 车牌定位, 边缘特性筛选, 数学形态学, 灰度纹理特征, 颜色特征, 多判定机制