Research On Visual-based Localization And Tracking Algorithm Of Humanoid Soccer Robot

The rapidly emerging robot soccer game is a high-tech competing activity with the de-velopment of intelligent robotics and distributed artificial intelligence; it has been one of the most challenging projects in the field of robotics and artificial intelligence research. The lo-calization and robot tracking are the basis of completing more advanced control decision in the soccer game. And it is difficult to localize and track robot in the environment of mul-ti-robots and multi-objects. With the development of computer vision, visual-based localiza-tion and tracking has been popular in recent years. Therefore, the research on visual-based localization and tracking of robot has important theoretical significance and research value.Based on analysis of the related technology and research of many literatures, this thesis focus on the research and realize the algorithm of robot localization and tracking. In the as-pect of robot localization, three aspects of research have been done:firstly, to strengthen the real-time and effectiveness of processing the information in the game, on the basis of building the color table, the visual-recognized area has been separated, the internal and external pa-rameters have been obtained through camera calibration, the monocular distance measurement model of robot Nao has been built, and this model is proved correct through experiments. Secondly, to correct the errors which are caused by the structure of Nao, the kinematic cali-bration is completed through building the D-H parameter model of Nao. Thirdly, for consid-ering the robot-kidnapping in the Robocup, robot Nao is localized through building the mo-tion model and sensor model of Nao and augmented MCL algorithm is used. In the aspect of robot tracking, three aspects of research have been done:firstly, robot is detected and recog-nized by threshold segmentation. Secondly, the applying of particle filter and Kalman filter in the object tracking has been researched and the effect of these two algorithms has been com-pared. Thirdly, robot tracking is completed by using particle filter.The above research obtained satisfactory results by a number of experiments. These re-sults not only illustrate the correction of theory or using the theory to solve problems of robot localization and tracking, but this kind of research is deserved to do, at the same time, it pro-vides the basis of multi-robots localization and multi-objects tracking research. This research can make it possible of completing more advanced game strategy.