足球机器人对彩球的鲁棒跟踪识别

Robust Tracking and Identification to the Color-ball for Soccer Robot

在基于视觉的足球机器人系统中,对场上焦点目标——球的动态跟踪识别是系统设计的第一要务。针对半自主微型机器人足球比赛中的小球易受场上干扰、小车遮挡造成的识别丢失问题,提出基于预测与搜索窗的图像目标跟踪识别方法。通过最小二乘法预测丢失小球的可能位置,将图像目标搜索限制在局部小区域内,并利用搜索窗内的在线状态信息加以判断,实现运动目标被遮挡情况下的有效跟踪识别。实验与比赛结果统计表明,该方法实时跟踪识别效果好、鲁棒性强。

The dynamic track and identification to the color-ball in soccer robot system based on vision is the most important task.In order to overcome the difficulty in identifying the ball sheltered by mobile robots in the robot soccer games,this paper presents a new visual tracking identification method based on an adaptive search window and the prediction to the location of the ball.The developed vision system has been successfully integrated into the soccer robot system.The experimental and practical results in robot soccer games show the method with good effectiveness and robustness.