基于形态学和surf特征的车牌定位算法研究

A License-plate Location Algorithm based on Morphology and Surf Features

提出一种基于车牌字符特征的车牌定位方法。该方法根据颜色及边缘信息通过数学形态学手段实现车牌区域初定位得到候选区域,再求取候选区域灰度图的surf特征并形成特征描述向量,然后利用最近邻匹配法找到正确匹配点对,最终根据候选区域与参考图像间的匹配度来完成精确定位,并同时求得变换矩阵实现所求车牌区域的倾斜校正。实验结果表明,该方法能准确、实时高效地实现车牌区域定位。

This paper proposes a license-plate location method based on features of the license plate. According to the color and edge information and by the methodology of mathematical morphology, the candidate area is acquired, and the initial rough location realized. And then the surf features of candidate regions are calculated and the feature vectors formed. Meanwhile, the method of nearest-neighbor matching is employed to find the correct matching points, and based on the matchability of the candidate area and the reference image, the exact location is finished, and the transformation matrix for license plate tilt correction acquired thus achieving the tilt correction of the wanted license-plate area. Experimental results show that the proposed method could, in real time, realize license-plate location accurately and efficiently.