Research And Design Of The Vision Sub-System In Small Size Soccer Robot

The Robot soccer games have been dramatically developed as a technical competition in recent years. It is a basic subject in the area of Manmade agent and Robotics ,and a program with many high-technique, it is very .It is a interdisciplinary field of different subjects, such as exact-mechanism, robotics, automated control, sensor and fusion, communication, computer vision and image processing, multi agent correspond and, wireless communication, etc. It is an ideal test bed of robotics and artificial intelligence. It is far-reaching academic significance and application that deeply research the Robot soccer games.This thesis focuses on The RoboCup Small Size League .The performance requirement and function of the vision subsystem is analyzed, and the design process of the system is brought forward. According to system functions, the system is divided into four modules: image collection, image segmentation, camera calibration and target identification. Based on traditional vision-processing methods, the performance is improved by using and improving the algorithm. Especially it is presented that a vision processing technique which using an improved color matching approach to build the object's color matching template, improved the recognition rate of the vision system for the RoboCup small size league under the variable lighting condition. The main research contents are as follows:(1) In image segmentation module, it is introduced that adopt an improved color matching approach to identify the color, thus the application of this algorithm has resulted in good effect under the variable lighting condition. (2) In camera calibration module, it adopts Tsai's two-stage technique and acquired the better effects.(3) In target identification module, it not only identified the robot's location, direction, number, but also forecasted the ball's position by using Kalman filter. The searching range is reduced, and the recognition rate of the vision system is improved. In the process of the target identification and data fusion, the concept of evaluation function is introduced, according to the value to distinguish target and this method can improve the accuracy.The system described in the thesis was integrated in the small-size robot soccer team DUT Fantasia SmallSize which participate RoboCup China Open 2007. In the games, the system worked well.