Research And Development Of Global Vision System In Robotic Soccer Game

With the development of cybernetics and computer science, robotic soccer hasbecome a research flat on which the theories of multi-agent system and relevantapplication study in distributed artificial intelligence have already become the studyfocus of artificial intelligence and intellectual control. In robotic soccer game ,themachine vision is the main approach through which the game environment isdetected and feedbacked to robots. It is very important for the team to win in thegame. the key of win. We have designed the machine vision subsystem of theAFU2004 RoboCup-Small team of robotic soccer of Southern Yangtze University(SYTU), and carried on relatively deep research work in following:I. The different designs of color blob are analysed and compared.The movingobject can be detected and traced in real time environment by machine visionsubsystem.II. RGB, YUV and HSV are compared. The faults caused by segmentation based ontwin-histogram algorithm are analysed. The segmentation method based onbackground-color-invariable is described in this paper, which can be used tominish the faults.III. The adpative segmentation method based on color edge detection and HSVthreshold map is developed by simulating the humanoid color constancy.IV. The Tsai's two-stage camera calibration technique is used in machine visionsubsystem. The geometric calibration method for lens by applying polynomialwarping is improved and the detective precision and efficiency are advanced.V. The geometric calibration method for lens by applying polynomial warping isimproved and detective precision and efficiency are acquired.VI. A formula is developed to correct the errors caused by projection.