The Research Of Machine Vision System And Its Application At Robocup

With the development of computer technology, especially the mature of multimedia technology and digital image process theory, rapid progress of large-scale integrate circuit; machine vision has been widely investigated by many countries. Owing to machine vision can acquire huge information in short time and easy to carry out automatism, integrated with design and machining control information, therefore, people use machine vision in modern automation process at various fields such as condition inspecting, finished product test and quality control etc.Robot world cup (Robocup) as an international competition provides us an ideal simulation and test platform for kinds of researches like machine vision, robots, artificial intelligence, multi agent system and pattern recognition etc. Machine vision system is an extremely important part of RoboCup competition, its performance directly influence the result of game. Because the position moving of robot players and the ball make the field has a protean condition, different period of time and zones of field has different illumination, and current robot vision system can't satisfy the request of precision and real-time orientation. So, how to design a vision system with environment adaptable and real-time high efficiency comes to an essential and key problem.After the investigating statuses in developments about machine vision in domestic and foreign fields are reviewed and summarized, the paper combines with the RoboCup rules and designs a robot vision system, meanwhile, discusses the topic of image process technology, color system and robot architecture which need by the vision system. Main work and innovation of this thesis:(1) First introduce the vision system of Robocup and its processing flow, and then design an integrated robot vision system and its pretreatment process. Finally, analysis and resolve the problem which maybe occur during the Robocup competition.(2) We proposed a fast simple geometric blob determination algorithm and an image search method based on position prediction. It can find the color mark's center on the top of robot player quickly and predict the start position of search process which increases processing speed and precision.(3) This paper use a color extract algorithm which combine RGB color space and HSI color space. Using RGB color model to search feature color, transform into HSI color model immediately as parameter of feature color when find the seed point. By this way, the vision system becomes more adaptable to various illuminations and enormously decreases the work of image analysis and processing.At last, we test the vision system by Robocup small team. Experiments show that the feasibility of soccer robot vision system satisfies the request of real-time, stability and precision and has a bright future in application prospects.