摘要
搭建了基于高速视觉的竞争型网络机器人系统.通过对比选取并融合了多种识别方法,设计了一套面向竞争型网络机器人的目标快速识别跟踪算法.采用动态窗口技术,并通过对不同传统方法的改进,进一步提高了该算法的实时性.本文方法能够在较复杂的背景下跟踪运动的目标机器人,具有较小的运算量及较好的鲁棒性.实验结果表明,该算法处理640×480像素的8位灰度图像的时间不超过5ms.
A competitive networked robot system based on high-speed vision is built.At the same time,by comparison, selection and integration of a variety of identification methods,a quick and effective target recognition and tracking algorithm for competitive networked robots is designed.Dynamic window technique and the improvement of traditional methods further enhance the system＇s real-time performance.The algorithm can track the moving target robot in complex background,with good robustness and less computation.Experimental results show that it takes less than 5 ms to process one frame of grey image of the size 640×480 pixels.
出处
《机器人》
EI
CSCD
北大核心
2011年第6期658-665,672,共9页
Robot
基金
国家自然科学基金资助项目(60905061
60875056)
天津市科技计划资助项目(08JCYBJC12700)
关键词
竞争型网络机器人系统
视觉伺服
高速视觉
目标识别
动态窗口
competitive networked robot system
visual servo
high-speed vision
object recognition
dynamic window
