The Research Of Large League Robot Soccer Vision System Based On Multi-Parallel

Robot Vision System, which is an important component of the robot system to achieve similar functions as the human eye, is a specific application of machine vision systems in robot system. The vision subsystem of Large League Mirosot, which was appeared firstly in FIRA 2004, is studied in this paper. And the Large League Mirosot vision system is designed to solve the problems led by the bigger play ground and more robots.First, we introduce the principle and component of vision system, so we can know how to construct the hardware system for the Large League Mirosot. Two cameras are used to meet the demand of view field and resolution. It also avoids the dead-zone problem. And a model based on multi-parallel has being adopted to realize image grabbing from the dual-camera. Result shows that the method satisfies the demand of image processing for the huge data amount.Second, in response to uneven distribution of illumination intensity, we study the color model theory, especially analyzed and compared the RGB, YUV and HSI color model. RGB color lookup table is constructed based on the HSI model. Color image segmentation is done based on this table. It's a solution to meet the needs of uneven distribution of light intensity and real-time identification.Third, an in-depth study of image distortion correction of wide-angle imaging system is conducted. An overview of the imaging camera model, distortion and distortion correction model is introduced. Based on this, a Newton iterative correction method is used to correct the distortion of the ground image. It improved the positioning accuracy of robots or ball. We also discuss real-time data communication, which is the core technology for multi-parallel. And also the information of robots or ball in the two half area grabbed by the two cameras is integrated.Fourth, a research of color patch is made for the Large League Mirosot robots. A circle color patch design and a rectangular color match design have been used in the vision system. And we analyze the merit and shortcoming for these methods. The experiments show that circle color match design has a greater accuracy of the center point of the robot than rectangular color. But in this method, 11 robots can be indicated with five kinds of color. On the contrary, the center point of the robot and the orientation angle are calculated with rectangular color match design, the accuracy may less than circle, but 11 robots can be indicated with only four kinds of color. According to the error allowed in the vision system, the rectangular color match design can achieve color design purposes well too.Finally, the visual system for human-computer interface is implemented with VC++ 6.0. And the RGB color lookup table based on HSI is structured. Polynomial image distortion correction, algorithm to identify the color match and real-time data communication are also implemented.