基于低秩纹理特征的车牌倾斜校正方法

Vehicle License Plate Tilt Correction Approach Based on Low-rank Textures

针对传统的基于直线检测的车牌倾斜校正方法对车牌边缘以及噪声较为敏感,并且不能同时检测车牌在水平和垂直方向的倾斜角度的问题,提出了一种基于低秩纹理特征的车牌倾斜校正方法。车牌图像的低秩先验信息是指对车牌图像,当车牌在图像中处于水平或垂直时,该图像对应的矩阵的秩是最小的。利用该低秩先验,采取一定的步长对车牌图像进行仿射变换,通过寻找最小秩对应的变换参数就可以获取车牌的倾斜参数,进而对车牌进行倾斜校正。实验结果表明,该方法能够准确快速地对车牌进行水平和垂直方向的倾斜校正,对车牌边缘、光照、噪声和分辨率等不敏感,具有一定的实用性。

The traditional license plate tilt correction methods were sensitive to plate edges and noise, and cannot gain the license plate tilt angles in the direction of the horizontal and vertical at the same time. To solve this problem, the paper proposed a license plate tilt correction method based on low-rank texture character. The low rank priori information of a license plate image is that the rank of the image is the smallest when the license plate in the image lies in a horizontal or a vertical state. According to the low rank priori, we can take a certain step size to operate the affine transform for the license plate image, and the tilt parameters can be gained by the corresponding transformation parameters which let the rank of transformation result be the minimum one, therefore we can use the method to correct the license plate tilt. The experimental results show that the method can accurately and fast correct the license plate tilt in horizontal and vertical direction, and is not sensitive to the license plate edge, light, noise and resolution, and then has a certain practical value.