干旱区生态环境知识资源中心

We have developed a mobile guard robot, which call autonomously get oil arid off the conventional elevators of a building for patrolling different floors. In this paper, we describe the vision system using a single camera oil the robot arid a single artificial landmark oil the wall to measure the 3D positions of elevator buttons, where we take advantages of two architectures of neural nets: one is the competitive architecture for pattern recognition invariant to projective transformations, and the other is the learning architecture for vector quantization in order to generate efficient template patterns. We have also developed a method to calculate the 3D positions using a single camera and a single artificial landmark for a simple, stable and low cost visual measurement system.