基于组合跟踪的景象匹配分析评估方法  

Analysis and Evaluation Method of Scene Matching Based on Combination Tracking

作　　者：涂国勇[1] 李伟建[1] 刘兴威[1] 刘旭光 李壮 谭凤义 TU Guoyong;LI Weijian;LIU Xingwei;LIU Xuguang;LI Zhuang;TAN Fengyi(Jiuquan Satellite Launch Center,Jiuquan Gansu,732750,China)

机构地区：[1]酒泉卫星发射中心,甘肃酒泉732750

出　　处：《海军航空大学学报》2022年第1期101-107,共7页Journal of Naval Aviation University

基　　金：国家自然科学基金(61402489)。

摘　　要：景象匹配飞行试验中,图像序列中存在的剧烈尺度变化及多种外界因素影响导致的几何变形等成像质量变化,增加了景象匹配全流程自动评估的难度。为了解决这个问题,提出1种基于组合跟踪的景象匹配图像全流程评估方法。首先,基于序列图关键帧选取原则,提出序列图关键帧匹配方法;然后将该方法与多种目标跟踪算法进行对比分析;再根据对比分析的结果,提出1种基于该方法和TLD组合跟踪的景象匹配全流程评估方法;最后,将组合跟踪方法应用到3组模拟实验中。评估实验结果表明,该组合跟踪方法能够可靠快速地进行序列图全流程评估,对图像旋转、尺度变化、图像噪声和模糊等都有很强的适应性,并能以二维关系图的方式给出景象匹配全流程的定量评估结论。In scene-matching flight trials,the drastic scale changes in the image sequence and the geometric deformation caused by the influence of various external factors increase the difficulty to evaluate the whole process of scene matching automatically.A whole-process evaluation method of scene matching based on combination tracking is proposed.Firstly,a sequence image keyframe method based on the principle of sequence images keyframe selection is proposed,and the method is compared with various target tracking algorithms.Then,according to the comparision analysis results,a scene matching whole-process evaluation method based on the combination tracking algorithm of sequence image matching and TLD is proposed.Finally,the combined tracking method is applied to three sets of simulation experiments.The results of evaluation experiments show that the method enables reliable and rapid whole-process evaluation of sequence images is highly adaptable to image rotation,scale change,image noise,image blurring,and gives quantitative evaluation conclusion in the form of twodimensional graphs.

关 键 词：景象匹配 全流程评估 组合跟踪 序列图匹配 匹配算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]