Abstract: A novel algorithm is proposed because there is high degree of distinction between navigation-line extracted by transformer substation robot and background area, and the images are distributed in regular ways.The algorithm first extracts the color component of navigation-line from the preprocessed images in the HSV space. Then, in order to quickly remove the background information to extract the navigation-line, the algorithm improves the Otsu algorithm by comparing and scanning the local variance and the maximal between-cluster variance of the image whereas the Cb component's grayscale distribution is insensitive to light intensity in the YCbCr space. The experiment results show that the overall efficiency of the algorithm is up to 98.9% in different weather and light conditions when it is applied to wheel robot based on the ROS system, and that the algorithm has high robustness.