基于定点航拍图像序列的运动目标跟踪方法  

A method to track moving target for aerial image sequence in combination with background stabilization

作　　者：郭清达[1] 武彦林[1] 于广平[2] GUO Qing-da;WU Yan-lin;YU Guang-ping(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510640,China;Shenyang Institute of Automation of Chinese Academy of Sciences at Guangzhou,Guangzhou 511458,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广州510640 [2]广州中国科学院沈阳自动化研究所分所,广州511458

出　　处：《信息技术》2017年第7期70-73,共4页Information Technology

基　　金：广东省协同创新项目产学研合作专项(2014B09090-1014)

摘　　要：定点航拍图像序列不仅包括目标物的运动信息,同时包括外界环境干扰或自身抖动信息。针对航拍图像序列的背景抖动,对前后帧进行Fast特征点检测匹配,并对当前帧进行仿射变换至前帧的参考坐标系中;针对航拍图像序列中的运动目标,采用Kalman滤波来预测运动信息,设置置信度区间减少Meanshift目标检测迭代的次数。仿真表明:结合Fast特征点的背景运动补偿与改进的Meanshift目标跟踪方法,可有效地补偿外界扰动或自身抖动对航拍图像序列的影响,并对目标物进行快速的跟踪。Fixed location aerial image sequences include not only the motion information of the target movement,but also the external or internal shaky information.To eliminate the shaky information in the background of aerial image sequence,this paper used fast method to detect and match feature points between adjacent frames,then transformed the current frame to the front frame in the selected coordinate system.Kalman filter was used to predict the motion information of target object.Meantime,the confidence interval was set up to reduce the number of iterations in the target object of in aerial image sequences.The computer simulation results show that the combination of feature points of background motion compensation and improvement of Meanshift target tracking method can effectively compensate external or internal shaky information and rapidly track the target object in aerial image sequences.

关 键 词：定点航拍 背景防抖 Fast特征点 Meanshift跟踪 KALMAN滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]