Research On Application Of Omni-directional Vision Based On Mobile Robot

Autonomous mobile robot is one of the most important topics on robotics research in recent years, which widely used in many applications, such as intelligent human-robot interaction, service robot, space robot and automatic vehicle. In these cases, it is an active field about mobile robot that how to observe the surrounding condition and track the target objects with the sensors carried on a mobile robot, so that makes the mobile robot more intelligent.This paper primarily does researches on omni-directional vision's basic theories based on mobile robot. And research work of application of omni-directional vision is done based on the basic theories. An omni-directional system is designed for mobile robot. With the development of the system, we focus on researches of omni-directional imagine unwrapping algorithm and target recognition and tracking based on omni-directional vision, the main content as follows:1. The system for omni-directional vision based on mobile robot is built, which mainly includes omni-directional vision sensor design method based on combination of MATLAB and VC++, hardware and software which are used for the vision processing. The system provides a good platform for future research work.2. Unwrapping algorithms of omni-directional imagine are studied in this paper, and several unwrapping algorithms are proposed. The fidelity unwrapping imagines and speed of the algorithms are compared according to the experiments.3. Interpolation algorithm based on omni-directional vision is proposed, which introduces interpolation algorithm into omni-directional unwrapping first. Firstly the basic interpolation algorithm of bilinear, cubic interpolation are introduced, then a new interpolation unwrapping algorithm is addressed according the analysis of omni-directional vision. The unwrapped imagine quality is improved effectively as the experiments shows. 4. Finally, the paper introduces target recognition algorithms of SSDA and Mean Shift into omni-directional vision processing. Two algorithms are improved to be applied to omni-directional vision. The effects of the algorithms are compared and experiments show that the improved algorithms meet with the application of omni-directional vision based on mobile robot, especially in capacity of real-time and right rate of recognition.