交互式多模型算法的MT-R机器人视觉跟踪  

Visual tracing algorithm based on interacting multiple model for MT-R robot

作　　者：弋英民[1] 黄莹[1] 

机构地区：[1]西安理工大学自动化与信息工程学院,西安710048

出　　处：《计算机应用研究》2014年第5期1574-1576,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(51275405);陕西省教育厅自然科学专项基金资助项目(2013Jk1078)

摘　　要：针对机器人的视觉跟踪问题,提出了一种基于交互式多模型算法的视觉跟踪方法。该方法采用HSV颜色模型,通过交互式多模型算法进行滤波估计,获得目标的运动属性。利用交互式多模型算法对目标的位置进行一步或者N步预测,在获得下一帧的观测值后,在预测位置的区域进行目标搜索,可减少搜索区域,节省搜索时间,增加了跟踪的实时性。使用该方法对机器人足球比赛中的红色球进行搜索跟踪。实验结果表明,应用该方法可获得目标的运动属性,能快速搜索到运动目标,并能进行准确跟踪。For the visual tracing problem, this paper presented a visual tracing method based on interacting multiple model （ IMM ） algorithm using HSV color model. It acquired the moving characteristics of the target by means of the filter function of IMM algorithm, and predicted the location of the target by IMM algorithm with one-step or N-steps. After obtaining the next observation frame, it searched the target in the predictive area. Thus in this way, it reduced the search area, saved the search time and increased the ability of real-time tracking. This method was put into practice to search and track the red ball in Robo- Cup. Experiment results show that the moving target can be searched quickly and tracked correetly. The moving characteristics of the target can be obtained with the proposed algorithm.

关 键 词：HSV颜色模型 交互式多模型算法 视觉跟踪 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]