In order to make the visually impaired people get the information of obstacles effectively and avoid it successfully in the unfamiliar environment, we propose an obstacle detection method based on depth information. Firstly, we use the edge characteristics of depth image to segment the obstacle by different depth. Then we remove the unnecessary ground information by gradient threshold. Our algorithm can label the obstacles by region growing algorithm respectively. Finally, we use rectangular windows to box out these obstacles. Our algorithm can display distances between Kinect sensor and these centers of obstacles on the frame for accuracy. Experimental results show that the proposed method has greater robustness than others. And the average of processing speed is only 0.08 second per frame.
Research Notes in Information Science (RNIS), pp.465-470