倒车辅助系统车位线识别方法研究

Parking Slot Marking Recognition Algorithm in Parking Assist System

提出了一种用于倒车辅助系统的基于改进Hough变换的车位线识别方法。算法首先采用局部像素分布特征以剔除非直线的干扰像素,然后使用像素直方图自适应地选择与车位线边缘直线相适宜的梯度方向区间,减少参与Hough变换的像素点数的同时减少背景干扰直线边缘的影响,最后通过判断相邻边缘间的像素灰度关系,进一步确认车位线边缘,从而识别出图像中的车位线。相比于传统hough变换和基于梯度方向区间的改进随机Hough变换(Grad-RHT),本文算法在保持与Grad-RHT运行效率相当的情况下,在各种干扰背景下都取得了更好的车位线边缘直线检测正确率,表现出较好的识别鲁棒性。

A new method based on improved Hough transform is proposed. Firstly, the local pixel distribution is used to eliminate the influence of outliers. Then, the suitable interval of the linear gradient to the edge of the parking line is adaptively selected based on the pixel histogram. Finally, the selected lines are identified by judging the gray relation,and thus the parking slot marking can be recognized. Compared with the traditional Hough transform and the improved random Hough transform based on the gradient direction region （Grad-RHT）,the proposed algorithm achieved a better recognition results and robustness under all kinds of interference background keeping the running efficiency of Grad-RHT.