Recognition of the unstructured road based on machine vision

Road recognition plays a key procedure in autonomous driving. Compared with laser radar, the method based on machine vision has the price advantage, also has a better prospects for development. Focusing on complex road surface and unclear boundary of unstructured road, a recognition method based on machine vision is proposed in this paper, which uses 2-d Tsallis entropy and double dogleg method. Firstly, the RGB images are converted to HSI color space. Then the hue and saturation images are segmented respectively, using the values calculated in the minimum 2-d Tsallis entropy. Considering the probability of white pixels in segmented hue image, the results are fused in different logic manners. Finally, the double dogleg method is used to recognize the edge of roads. Experimented with different material road images, this method is also proved to be effective.