基于边缘编组的车位线检测方法

针对自动泊车系统中检测车位线的问题，提出基于边缘编组的车位线检测方法。首先对图像进行预处理，并对传统Canny边缘检测算子改进，增加45°和135°方向模板计算梯度幅值和方向，降低噪声干扰并提高图像质量；接着，对边缘图像进行8邻域搜索并聚类分组，剔除较短直线、解决过连接问题；最后利用各边缘组内相邻两像素点梯度方向差分剔除非车位线特征的边缘组，通过对每个边缘组做随机Hough变换，并利用先验知识和最小二乘法拟合得到车位线。实验结果表明，该算法相对于传统Hough变换可以在不同情形下准确检测到车位线，并且具有较好的实时性。

Research on parking line detection method based on edge grouping

Aiming at the problem of detecting parking space lines in automatic parking systems, a parking space line detection method based on edge formation is proposed. Firstly, the image is preprocessed, and the traditional Canny edge detection operator is improved. The 45° and 135° direction templates are added to calculate the gradient amplitude and direction to reduce noise interference and improve image quality; then, the edge image is searched for 8 neighborhoods and Clustering and grouping, removing shorter straight lines, and solving the connection problem; finally, using the gradient direction difference of two adjacent pixels in each edge group to remove the edge groups that are not characteristic of the parking space line, by doing random Hough transform for each edge group, and using Prior knowledge and least squares fitting method get the parking space line. The experimental results show that compared with the traditional Hough transform, the algorithm can accurately detect the parking line in different situations and has better real-time performance.