Autonomous Visual Navigation Guided by Path Boundaries for Mobile Robot

A visual navigation method guided by path boundaries is presented, boundaries are detected and located by visual technologies, and the controls parameters are obtained according to boundaries for autonomous navigation. For the detection of boundaries, an image segmentation method based on sub-region growing is presented. In this method, the seed regions are found by the growing of sub-regions, and the pixels of these seed regions are segmented. The seeded procedure combines with segmentation, and the efficiency of segmentation is increased. In the course of the locating of boundaries, the located precision is related to the position of the targets when using planar homography estimation. Focus on this problem, a visual locating method using planar homography estimation based on improved RBF network is proposed. RBF network fits the relationship between the image plane and the path plane by its non-linear approach characteristic, thus the locating precision is improved. Experiments verify the effectiveness of the presented methods.