基于像素局部对比度的直线特征检测算法*

本文提出了基于像素局部对比度和全局虚警抑制的直线段检测算法。本文算法分为四步：第一步,计算像素点的梯度以及局部对比度；第二步,将方向大致相同的像素点分成直线支撑区域(像素方向为像素梯度方向的顺时针垂直方向),并对支撑区域进行矩形近似；第三步,在直线支撑区域内拟合直线段,计算直线参数；第四步,根据直线支撑区域信息验证直线段,抑制虚警。实验表明本文算法在复杂条件下,特别是光照影响条件下,具有更好的鲁棒性和效率。

Linear Feature Detection Algorithm Basedon Pixel Local Contrast

This paper proposed a line segment detector based on local contrast of pixels and global false detection control. Our method consists of four steps. Firstly, the gradient and local contrast of each pixel is calculated. Secondly, the pixels with the similar orientation are clustered into the LSR (Line Support Region) which is approximated by a rectangle. After that the LSR is fitted into a line segment with above one pixel width and the line parameters are estimated. Finally, the line segment candidates are validated and the false detection is controlled. The experimental results show that the proposed method is much more robust and efficient than the traditional methods in hash conditions, especially in complex illumination.