基于因子缩放的车牌图像倾斜校正方法

Approach for VLP Image Tilt Correction Based on Factor Zoom

针对倾斜的车牌图像对后继字符切割带来不利影响的问题,提出一种车牌图像上"字符"的宽度按高度同因子缩放的车牌图像倾斜校正方法。首先,根据车牌图像的特点将车牌图像分成四块,并使用Radon变换分别检测分割后的四块车牌图像;然后,根据检测的结果,求出车牌图像上车牌四条边框所在的直线方程以及对应的长度,并计算出上边框与下边框对应点连线的直线方程和长度,根据这个长度计算出连线上各点处的缩放因子;最后,根据所求的因子对倾斜车牌图像上所有的点进行校正。实验表明,与经典校正法相比,本文所提出的车牌图像倾斜校正方法校正的结果更符合实际的长宽字符比,校正效果更好。

In accordance with the problem that the inclined Vehicle License Plate（VLP） image brings adverse infection for the subsequent steps of character segmentation,a VLP image correction approach based on factor zoom which is computed by the scale of character height is proposed.Firstly,according to the features of the VLP image,the VLP image is divided into four parts,and every part is detected by Radon transform.Then four equations and lengths of frame in the VLP image are solved,and the lengths,equations and ratios of the corresponding dots of the top frame and the bottom frame are computed.Finally,every dot in the tilt VLP image is corrected according to the proportion.Experiments show that the proposed VLP image correction approach can obtain better ratio than the classic correction method.