足球机器人视觉系统中彩色目标分割方法

来源期刊：中南大学学报(自然科学版)2002年第4期

论文作者：徐大宏 吴敏 曹卫华

文章页码：428 - 430

关键词：足球机器人;彩色视觉;图像识别;二值化;全局扫描

Key words：soccer robot; color vision; image recognize; image-binary; whole scan

摘    要：足球机器人(MiroSot系列)仅依赖于其视觉系统获得比赛场上的信息.视觉系统对获取的小车顶部标志块的彩色图像进行处理,以识别小车信息.作者试验了3种彩色标志块的设置,并在第1种方案的基础上研究了一种快速彩色目标分割方法,即根据足球机器人视觉系统所处理的视频图像中,待识别的目标区域(前景)面积远小于其他区域(背景)的面积,运用一种基于全局扫描的快速算法,1次分割出所有感兴趣的目标,缩减了系统中目标识别时间,从而使系统能进行一些额外的图像预处理操作,在提高系统识别速度的同时,提高识别精度.

Abstract: Soccer robot(MiroSot) only depends on its vision system to get the information on the playground. In the vision system the color image of sign region on the top of robot is processed and analyzed to recognize robot. In this paper three schemes of color sign region are introduced. Based on the first scheme, a fast color object segmentation is also introduced. The area of foreground is smaller than that of background of the video image processed in the vision system of soccer robot. A fast algorithmbased on whole scan is introduced to find all the related objects by scanningonce, so the time of object recognition can be greatly cut down, and the system is able to do some additional image pre-processing to improve the precision of object recognition and the speed of object recognition.