一种车牌快速矫正算法

A Fast Algorithm for License Plate Recognition Correction

针对车牌识别中的图片倾斜矫正,提出了一种基于字符形态学特征,只需处理字符与字符之间的关系的新算法。该算法通过枚举扫描线的倾角代替对整个图像的旋转,节省了对图像进行旋转操作的时间。与传统的Hough变换、Radon变换以及近几年新兴的车牌矫正算法等方法相比较,该算法在保持高正确率的前提下,整体运行时间提高了3~5倍,减少了因车牌边框不明显而引起的误差,降低了水平矫正算法复杂度。实验结果表明,该算法效果良好。

The paper proposes a new algorithm based on the morphological characteristics of the character for license plate recognition image correction.The algorithm only processes the relationship between characters and characters.Instead of rotating the entire image,this algorithm enumerates scanning line＇s angle,which saves the time to rotate the entire image.Compared with the traditional Hough transform,Radon transform and new license plate correction algorithm in recent years,this new algorithm has its efficiency improved 3—5 times with higher accuracy and reduces the chances of errors caused by obvious license plate frame and the complexity of horizontal correction algorithm.The experimental results show that the algorithm works well.