足球机器人系统中一种快速模式识别算法的研究与实现

Research and Implementation of A Fast Algorithm for Pattern Recognition in Robot Soccer System

机器人足球赛是一项非常复杂的高技术对抗活动 ,快速准确的识别出机器人和小球是决策和控制的依据 .提出了一种高效准确的模式识别算法 ,该算法基于球场中各个机器人外罩顶部的色标颜色 ,首先进行网格扫描搜索切入点 ,再采用基于彩色图像的轮廓提取与跟踪 ,得出各个色块的轮廓 ,然后由预先设定的阈值来过滤无效色块 ,最后根据标定的机器人色彩 ,进行有效色块的最佳匹配 ,从而确定出球场上机器人的位置和方向以及小球的位置等信息 .实验表明 ,该方法能满足实时性要求 ,并能自适应多种亮度环境 .

Vision subsystem is an important component in robot soccer system that is required to be fast, accurate and robust against noise.An efficient and accurate algorithm of pattern recognition is presented. The steps of this approach can be summarized as follows: firstly, according to the size and shape of the two color markers on top of each robot in the playground, grid scan is adopted in order to find the first point for contour tracing; secondly, all the possible image blocks are picked up by contour tracing based on colorful image; thirdly, some image blocks are considered to be noise image blocks or repeated image blocks and are to be filtered according to the specified threshold; finally, in according to the given color, best match of the valid image blocks is to be done. Thereby, the position and the orientation of the robots, as well as the position of the ball are to be received. Several experiments prove that this method can self adaptto variable luminance and it can meet the requirements of real time calculation.