Color Image Stereo Correspondence Algorithm for Obstacle Detection

Obstacle detection is a crucial problem of autonomous vehicle navigation. The paper presents color image stereo correspondence algorithm for obstacle detection. Firstly, the color images are segmented in HSV color space. And the regions of interest are extracted from the images; Then, the rectangular boundary of the interest regions are calculated; Obstacle position finally are obtained using stereo correspondence method. Segmenting and extracting of obstacle object is the key during this process. The image in HSV color space in the paper has compact information structure. And every component of HSV color space is more independence. Image segment is more accurately with these characteristics. The result of experiment indicates that it is effect for detection of road, vehicles and house and suppression of natural background.