基于最小二乘和最小投影距离的车牌倾斜校正

在车牌图像的采集过程中,经常会有车牌倾斜的现象发生,这种倾斜给后续的字符分割和字符识别造成了很多不利影响。为此,文中提出了一种基于最小二乘和最小投影距离的车牌倾斜校正方法。该方法将车牌倾斜分成水平倾斜和垂直倾斜两部分:对于水平倾斜,首先对二值化后的车牌去边框和铆钉,再对车牌利用最小二乘拟合直线求取倾斜角;而对于垂直倾斜,则引入分块查找法来降低查找最小投影距离的执行次数,从而提高算法的执行效率。实验结果表明:该算法简单实用,能够准确地对车牌进行校正。

A slant correction algorithm for license plate based on least squares and minimum projection distance

During the acquisition of license plate image, image slant happened frequently makes the subsequent tasks of character segmentation and character recognition difficult. A new kind of slant correction algorithm based on least squares and minimum projection distance is presented. The algorithm proposed decomposes the license plate slant into horizontal slant and vertical slant. For horizontal slant correction, the slant angle is derived by fitting a straight line with least squares constraint after the removal of the border and rivets from the binarizated license plate image, while for vertical slant correction, execution times of finding the minimum projection distance are lowered by the introduction of block search method, hence the improvement of the algorithm execution efficiency. Finally, the experiment results show that this algorithm is simple and practical, and the slant of license plate can be corrected using the new algorithm.