Research And Design On A Vision Sysytem Of Centralized Control Soccer Robot

Computer vision has been more and more applied to all kinds of areas of life,along with the continuous innovation and improvement of computer vision algorithms,and people demand more and more from the computer vision,which requires that the innovation of computer vision can not be limited to theoretical research and needs computer vision to keep developing in reality.Soccer robots' vision system is a typical example of the application of computer vision,which makes full use of image acquisition and segmentation,feature extraction and target detection and tracking.This vision system provides a good experimental platform for the computer vision applications.This paper designs a vision system of soccer robots,according to the rules of FIRA's MIROSOT-3V3 game,which is one of the most influential soccer robots' tour matches in the world.The main contents include follows:(1)According to the environment of MIROSOT-3V3 game the camera,lens and image acquisition card are selected,and the hardware environment of the MIROSOT-3V3's vision system is set up.(2)A colorful image segmentation method based on a color look-up table is designed,which is suitable for MIROSOT-3V3 game.According to the change of targets' color,a method of selection of colorful threshold is presented.The above algorithms are achieved to verify those feasibilities using OPENCV,C++,MFC.(3)According to the rules of MIROSOT-3V3 game,color codes on the soccer robots are designed.A grid search method which is improved to locate targets preliminarily is given.A method to distinguish and implant seed points among targets is proposed.A fast seed points expansion method to extract the targets' features is established.A fast method is applied to judge position and direction of soccer robots and its feasibility is verified by programing softwares.(4)A tracking algorithm which is suitable for the MIROSOT-3V3 game and a plan to achieve the above algorithm are designed.