The Study Of Autonomous Mobile Robot's Vision System

This dissertation researches the computer vision of a four-wheeled autonomous mobile robot, it contains mainly three parts.In the first part, a global location method based on circular beacon is proposed. With imaging model, a space circle is reconstructed after quick recognition of the ellipse in the image. The mobile robot is located by combining the result with prior knowledge. This location method based on vision provides higher precision.The second part provides a novel multi-objects tracking algorithm with YUV color space as character vector space and gives the application in global vision. The self-adaptive color template overcomes the defects of traditional color identification algorithm whose performance is affected greatly by diverse light conditions. The idea that combines position prediction and color identification solves the problem that multi-objects have identical color. The tracking process is accelerated by spiral search and regional growth method.The third part presents a line-tracking method dependent on computer vision and fuzzy control. Since there are many lines in indoor environment, we give a quick recognition of target line. After deducing the kinematical model of the four-wheeled mobile robot, a fuzzy controller, which can imitate human in driving, is designed to control the mobile robot to track the line. Finally, the result shows that the algorithm is effective in line-tracking.