Intelligent Reconfigurable Mobile Robot Vision System For Study

The elicitation of the human physical and intelligent system is very significant to the navigation control of mobile robots under unknown environments. Visual information is one of the most important sources of environments' information for human, and the research of robot vision cannot be ignored. Because of the complexity of human's vision system and the lack of cognition about it, robot vision system is still faced with many difficulties. The watching and recognition of human are working as an integrated system, which combine with capturing visual information, recognizing environment, knowledge reasoning, fusing information, learning and so on. Though significant advances have been made in such fields, researchers have not paid enough attention to study the whole frame of integrated system until now. Therefore the thesis just focuses on the integrated architecture of system with generality, validity and flexibility. In this thesis, both the designs of software and hardware of the system have been discussed in details and some experiments have also been given.Firstly, the hardware design of the robot vision system based on multi-camera mode has been discussed. Considering the limitation of the normal computer vision, the hardware design can satisfy the different situations for mobile robot navigation. Parameters of the system have been analyzed in details. As a result, the robot can obtain much visual information synchronously.Secondly, the intelligent and reintegrated software architecture (IRSA) of mobile robot vision system based on agent technology has been presented. Focusing on the limitation of Marr vision theories and uniting in purpose-active vision theories, agent has been used as the basic model in the IRSA. The architecture can realize an intelligent software architecture with generality, validity and flexibility.Finally, an application of the IRSA has been given. A software system has been developed, which is using single-camera to trace some color and to judge fuzzy direction of the robot. The results show that our designs are valid and efficacious in dealing with robot vision issues.