摘要： 导航线的宽度、灰度在图像中具有均匀一致性，根据这些特征，可以进行导航线的鲁棒检测。对图像上的一行，先进行水平方向上的边缘检测，对相邻的2个边缘点，如果其距离大于导航线宽度减2，而小于导航线宽度的2倍，并且这2点间像素灰度值的方差比较小，则这2点就可能是导航线的边界点。对图像的若干行进行处理得到这些点对后，利用快速哈夫变换去除非导航线边界点，再利用最小二乘法精确拟合得到导航线的直线方程。针对实际图像的检测实验表明，该算法能从复杂背景中快速鲁棒地检测出导航线。

关键词: 导航, 直线, 检测, 视觉

Abstract: The width and the gray level of the navigation line keep almost the same in the image. Based on those characters, the navigation line can be detected robustly. Scanning the image alone the horizontal line to find the neighbor edge points whose distance is between the navigation line width minus 2 and the navigation line width multiple 2, and the gray level variance of pixels between them is small. Then the navigation line can be detected from those points by fast Hough transform roughly and refined by least square method. Experiments with real images show that this algorithm can detect the navigation line robustly with high speed.

Key words: Navigation, Line, Detection, Vision