Panoramic Vision Football Robot Targeting System

The study of panoramic vision targeting systems is crucial in the Soccer Robot Competition. It is the basis and foundation for the autonomous robot to achieve independent movements and make behavioral decisions. The study of the system involves many areas, so it needs a strong engineering background, and there is an important research value in this subject.In this thesis, the study of autonomous soccer robot is based on the Beijing Bo Chuang Voyager-Ⅱsystem platform and also the systematic study of the autonomous soccer robot target tracking positioning platform. This target tracking positioning platform is based on the Mean Shift Algorithm.Firstly, this thesis studies the robot's panoramic images based on the rules of Robocup Competition, including how to preprocess images and how to corrode and expand the acquired images for obtaining the best target detection results.Secondly, the system explores a method used to process images in terms of moving target detection, determining the method of processing the target image by contrasting the RGB, YUV, HSI. The system also gives a specific conversion formula, thus providing the basis for converting the RGB images collected from Panoramic Robot Vision to HSI images which have strong adaptability to changing light.Thirdly, this thesis studies the Mean Shift Algorithm in detail, and then proposes an adaptive tracking window width to improve the tracking accuracy. This Mean Shift Algorithm is combined with the Kalman filter algorithm to solve occlusion, thus achieving improved tracking. Real-time tracking of the orange football using this combination minimizes defects and increases the accuracy of target tracking.In summation, further research is needed to integrate panoramic vision, Mean Shift Algorithm, and the Kalman filter into the context of the Robocup Robot Soccer Competition.