基于图像特征的运动目标识别与伺服跟踪  

Recognition and servo tracking of moving target based on image features

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,北京100083

出　　处：《仪器仪表学报》2008年第3期644-648,共5页Chinese Journal of Scientific Instrument

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

摘　　要：本文研究了一种在室内相对复杂背景下机器人的运动目标识别与跟踪方法。基于Intel的计算机视觉库OpenCV进行图像处理与特征提取,采用图像的区域特征———图像矩作为特征信息,在有干扰的情况下,实现运动目标的形状识别和基于特征的视觉伺服跟踪控制。为提高系统性能,建立曲线拟合模型来预测运动目标的动态趋势和特征。实验结果表明,系统在不需要摄像机精确标定、背景有干扰的情况下,能够实现对运动目标进行实时稳定快速的识别和跟踪。In this paper, the research on the recognition and servo tracking of moving target for a wheeled moving robot with a monocular camera is carried out in a relatively complex background. At first the computer vision library from Intel, Open CV, was employed to implement image processing and feature extracting, then the shape of moving target was recognized and the visual servo tracking was achieved based on local features of image moments in noise condition. In order to improve the system performance, a curve-fitting model was built to predict the dynamic trend and characteristics. Experiment results indicate that the system can stably recognize a fast motional target and track it in real-time under noisy background and without precise camera calibration.

关 键 词：运动目标识别与跟踪 形状特征 特征视觉伺服 加窗滤波预测 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]