Construction Of RoboCup Small Size Soccer Robot Vision System

The competition of Soccer Robot is a standard problem for the studying and the education of the AI and robot technology after the "Deep Blue". RoboCup Small Size Leage is one of best in the RoboCup Worldwide Competition. It focus on intelligent multi-agent cooperation and control in a highly dynamic environment with a hybrid centralized/distributed system related with the vision, wireless communication, real-time program, strategy control and so on. The vision subsystem is the front-end of the soccer robot and the application of the machine vision in soccer robot. Its key task is to get the information about the outer circumstance and to recognize, relocate and trace the target in the field. There are many things can give the design of the vision system so much challenge, such as the nonuniform and instability of the light, the high speed of the target, the requirement of high precision, the calibration of the camera lens and so on.We focus on the construction of the RoboCup Small Size vision subsystem, emphasize introduce the design scheme, and have a detailed introduction of the vision process system software RoboCup SSLImageView platform. According to the characteristic of small size vision subsystem, we analyze the hardware environment, hardware selection, the design of the hardware port and the design of the image process software. The image process software can be divided to three modules - image segmentation, camera calibration and object recognition for the analysis of the functions of the vision subsystem which will be showed in detail individually. The following is my research achievement:1) Design and develop the RoboCup SSLImageView image process software, and gain the computer software composition right of RoboCup Small Size Robot Soccer vision system software.2) Design an auto threshold method which is high efficient and precise and has the ability of dealing with the complex field situation for the drawback of the manual threshold method.3) Bring forward a threshold auto-adapting based in the field background idea which can work in the unstable situation.4) Design improved edge searching method which advances image process speed.ZJU. RoboCup Small Size Team had a new breakthrough in 2006. We firstly got into the one of the best eight in RoboCup2006 Bremen Germany, and got the national champion in ChinaOpen 2006. The RoboCup SSLImageView vision system software represents the great stability and good efficiency.