光学卫星影像配准同名点快速提取方法  

A rapid extraction method of tie points from optical satellite images

作　　者：赵鑫 王萍[1] 李慧 荆林海 赵晓晴 ZHAO Xin;WANG Ping;LI Hui;JING Linhai;ZHAO Xiaoqing(College of Geomatics,Shandong University of Science and Technology,Qingdao,Shandong 266590,China;Key Laboratory of Digital Earth Science*Aerospace Information Research institute,Chinese Academy of Sciences,Beijing 100094,China)

机构地区：[1]山东科技大学测绘与空间信息学院,山东青岛266590 [2]中国科学院空天信息创新研究院数字地球重点实验室,北京100094

出　　处：《测绘科学》2021年第10期98-107,共10页Science of Surveying and Mapping

基　　金：国家重点研发计划重点专项课题项目(2017YFC1500902);国家自然科学基金青年基金项目(41801259);海南省自然科学基金面上项目(418MS113);新疆自治区重大科技专项(2018A03004)。

摘　　要：针对多源高分辨率遥感图像控制点提取时存在错误匹配点多、分布不均匀、大幅面图像特征提取效率低等问题,该文基于传统的尺度不变特征变换(SIFT)算法提出了一种改进的高分辨率遥感图像配准同名点快速提取方法。该方法首先将待配准图像按网格分块为子图像,并基于地理信息约束得到各子图像对应参考子图像;然后将ShiTomasi角点和SIFT描述子结合,在每一对子图像上进行特征点的提取与特征匹配;再利用随机采样一致性(RANSAC)算法和最小二乘迭代法剔除错误匹配点,并基于贪心算法剔除冗余的控制点,最终得到分布均匀的配准同名点。利用平原和山区两组典型多源遥感图像进行了实验,并利用提取的同名点进行图像配准,结果表明,跟传统SIFT方法、采用分区策略的SIFT方法相比,该文算法在同名点数量、匹配点对分布的均匀程度、匹配速度以及配准精度上都有较大的提高,能满足大幅面图像配准的需求。Aiming to the problems of the control points extraction from multi-source high-resolution remote sensing images,such as a large number of error tie points,uneven distribution,and low efficiency when dealing with large-size images,an improved tie points extraction method,based on traditional scale-invariant feature transform(SIFT)algorithm,for multi-source high resolution remote sensing image was proposed in this paper.Firstly,the warp image was divided into several sub-images according to the uniform grid,and the corresponding reference sub-images of these sub-images were obtained based on the constraint of geographic information.Then,the Shi_Tomasi corner and SIFT descriptor were combined to extract and match tie points generated from each pair of the sub-images.After that,the random sample consensus(RANSAC)algorithm and the least square iteration algorithm were employed to remove error tie points,and the redundant tie points were then eliminated by the greedy method of the maximum clique problem.Finally,the evenly distributed matching points were obtained.The proposed method was compared with the traditional SIFT method and the SIFT method using the sub-image approach,using two groups of multisource high-resolution remote sensing images covering plain area and mountainous area,respectively.The experimental results showed that the proposed method had significant improvements in the number of tie points,the accuracy of registration,the uniformity of distribution,and the matching efficiency,indicating that the proposed method could meet the requirements of large-size image registration.

关 键 词：多源高分辨率遥感图像 同名点提取 Shi_Tomasi角点检测 SIFT 

分 类 号：P237[天文地球—摄影测量与遥感] TP751.1[天文地球—测绘科学与技术]