An integrated approach to vision-based autonomous land vehicle (ALV) guidance for automatic navigation and collision avoidance in building corridors and elevators is proposed. Computer vision techniques are utilized to locate an ALV in building corridors and to control it to enter elevators by the use of multiple corner points on walls. The location principle is based on matching input corner position information with a model using a distance-weighted correlation function. Futhermore, radio equipment is employed to control the elevator operations of lifting up, lifting down, door closing, and door opening so that the ALV can enter elevators automatically. Finally, a photoelectric sensor-based obstacle avoidance system for the ALV is developed. A real ALV was constructed as a testbed. Many successful navigation experiments confirm the effectiveness of the proposed approach.
Proceedings of 1992 Conference on Computer Vision, Graphics and Image Processing, Nantou, Taiwan, Republic of China,