This paper studies an algorithm that navigates a robot in an unknown environment with a stereo vision system. Different from a commonly used SLAM method that computes the distance between the robot and a landmark, our study intends to autonomously find an opening, such as a hallway or a window, for the robot to pass. After transforming the regular RGB images, we perform the 3D scene reconstruction in the gray-scale space. The sum of squared differences (SSD) algorithm is applied to match pixels and estimate the shifting distances. An algorithm is proposed to determine the passageway of the environment for guidance purpose. Experiments are presented to demonstrate the validity of our study, and compared with the commercial Kinect kit. The result of the research is especially applicable to surveil in an unknown building or an indoor disaster site.
Relation:
Journal of Aeronautics Astronautics and Aviation 52(4), p.361-370
