A Computer Vision For Navigation Of Mobile Robots

摘要

A computer system which can find the pathway in the robot's sensor field of view with only one camera has been developed. The low level extraction of linear segmentations is performed by edge following algorithm with the local gray level and gradient as heuristics. The symbolic lists which describe the optical and geometric properties of linear segmentations are generated as the output of the low level and submit to the high level. In high level stage a reasoning mechanism whose knowledge base consists of the rules modeling three basic types of pathway segments (non-branch, branch, and end) works on the symbolic lists to make the judgement of which type to be happened in the sensor area. Several parameters such as the exact positions of the pathway can be calculated as the signals for navigation control system. The experiments showed that the system behaved specially well for the existence of noise. shadow. and irregularities of environments.