直线检测的灰度投影积分方法

Gray projecting integral method for line detection

为了提高图像上直线检测的鲁棒性和精度,提出一种灰度投影积分(GP I)方法。GP I方法将需要检测的图像直接在直角坐标空间沿某个投影方向进行投影,将对应到同一投影点的像素灰度值进行累加,得到灰度投影积分向量,再旋转投影方向计算不同投影方向的灰度投影积分向量,从而得到一个原图像的灰度投影积分矩阵,再由该矩阵的极值元素所在的行和列求出图像中直线的方程。实验结果表明,GP I方法不需要事先提取边缘点,就能够快速准确地检测出图像中的直线,避免了传统方法在提取边缘点时阈值选取不当带来的直线检测误差。

A gray projecting integral （GPI） method was developed which improves the robustness and accuracy of line detection in images. The method projects the image to a plane in Cartesian coordinates with the gray values of pixels projected to the same point added to produce a GPI vector. The projecting direction is then rotated to compute various GPI vectors for a GPI matrix of the source image. The equation for the line in the image is then obtained from the row and column indexes of the extremums elements in the matrix. Experimental results show that the GPI method can quickly and accurately detect the line in point detection, which avoids the image without previous edge errors resulting from inproper thresholds in the edge point detection when using conventional methods.