A vision-guided object tracking and prediction algorithm for soccer robots

摘要

This paper presents a fast and effective visual tracking and location prediction algorithm for cooperative multiple soccer robots playing a soccer game. The picture image is captured by the system and the motion parameters of each object are then computed from the captured image. Then, an image is selected as a template and 2D color image pattern is extracted from the incoming image and a sequence of images are generated by using a simple linear predictor that speeds up the processing time. Based on the computed tracking parameters, a heuristic path is generated. By using a vision-based command to line-of-sight (CLOS) guidance method that predicts and generates next robot motions. Monitoring the selected object using visual images, the vision system tracks the object and adjusts the velocities and directions of the robots to cope with sudden changes of direction such as bounded ball from an obstacle or expected position errors of the selected object. The simulation results as well as experimental results show that the presented algorithm works well for the soccer game.