双目轮式移动机器人的运动目标识别与跟踪  

Recognition and tracking of moving targets for binocular wheel moving robotst

作　　者：刘宏鼎[1] 秦世引[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京市100083

出　　处：《智能系统学报》2007年第3期19-24,共6页CAAI Transactions on Intelligent Systems

基　　金：国家高技术研究发展计划资助项目(2006AA04Z207);教育部博士点基金资助项目(20060006018)

摘　　要：研究在室内相对复杂背景下的运动目标识别与跟踪,采用Hu氏不变矩作为目标特征,通过环形方式搜索种子点进行目标区域生长,在有干扰的情况下检出目标,并估计相对位姿和控制本体机器人快速跟踪目标.用Intel的计算机视觉库OpenCV实现图像预处理,图像分割,孔洞填充,区域生长和特征提取.实验结果表明,系统能对运动目标进行实时稳定快速的跟踪,适用性强.This paper studies in depth recognition and tracking of moving targets by binocular mobile robots in relatively complex environments. To deal with noise, Hu＇s moment invariants were taken as object features and object region growing was carried out from the seed point which was searched by loop scanning. In this way objects could be detected and their relative positions estimated so as to effectively track moving targets and control the individual robots. This paper made use of OpenCV to carry out image preprocessing, image segmentation, hole filling, region growing and feature extracting. Experimental results indicated tha the system produces stable and fast tracking of moving targets in real time and is suitable for practical applications.

关 键 词：运动目标 识别与跟踪 区域生长 形状特征提取 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]