利用投影最小距离的车牌校正优化算法

An Optimized Algorithm of Vehicle License Plate Correction Using Projection Minimum Distance

提出一种利用字符投影最小距离进行的车牌校正方法。该方法根据车牌垂直旋转计算字符区域在垂直坐标轴上的投影,利用投影的最小距离获取垂直倾斜角度;根据车牌水平错切计算字符区域在水平方向投影,利用投影最小距离获得水平错切角度;最后进行仿射变换,并利用双线性插值对图像进行校正。校正过程中,采用折半查找和记录校正角度,最后利用原始图像校正一次获得结果图像来优化算法,以减少计算次数和毛刺的产生。该方法不依赖车牌边框特征,抗干扰性强。实测车牌图像证明,该方法具有较高的鲁棒性和工程实用性。

This paper presents a new method based on VLP character projection axis minimum distance. The method shows that the VLP character area is projected to vertical axis by rotating the VLP binary image and the tilt angle is captured by vertical projection minimum distance. The character area is projected to horizontal axis by sheering the VLP binary image and the sheer angle is captured by horizontal projection minimum distance. At last the corrected image can be executed by Affine transform and bilinear interpolation. For the purpose of optimizing the correction algorithm, binary search and recording the correction angle are used to improve the processing speed and to reduce the burrs of result image. The method with good anti-interference does not depend on the VLP frame feature. The experiments on actual images that were taken under various conditions from traffic stations show good robustness and practicality.