摘要
本文提出了一种基于控制点库的SIFT多源大幅遥感影像自动配准方法,该方法首先设计并建立控制点影像库;然后采用地理坐标粗定位及SIFT算法自动精确查找同名控制点;利用影像分块迭代最小二乘法拟合,去除错同名控制点;最后运用多项式模型完成影像配准。以环境减灾小卫星CCD及近红外、QuickBird 2.4m及0.6m、TM及印度星(irs-p6)、LiDAR强度图等影像为实验对象。结果表明该方法能有效处理大幅遥感影像,同时针对多源遥感影像其配准精度达到亚像素级(RMS 0.617～0.934)。
The automatic image registration based on GCP database with SIFF algorithm was studied in this paper. The method designed and builded a GCP database to store information of GCP images firstly. Then the algorithm approximately located the GCP images in the warp images by geographic location. SI^l~ algorithm was used to locate the position of the GCP images accurately. After all the GCP points were obtained, the algorithm used least square method to reduce the wrong GCP points. Finally, the polynomial model and the correct GCP points were used to warp the images. Large remote sensing images （HJ-1A ／ B, 875M） and Quick Bird ／ TM ／ LiDAR etc. images were used to test the method. It indicated that the method could process the large images effectively and the precision of warped image was sub pixel （RMS 0. 617 -0. 934） .
出处
《测绘科学》
CSCD
北大核心
2011年第4期35-38,共4页
Science of Surveying and Mapping
基金
国家重点基础研究发展计划项目(2007CB714406)
国家科技支撑计划(2008BAC34B03)
中国科学院知识创新工程青年人才领域前沿项目专项项目资助
中国科学院遥感应用研究所遥感科学国家重点实验室资助项目
欧盟项目CEOP-AEGIS(FP7-ENV-2007-1 Grant nr.212921)
关键词
影像配准
自动配准
GCP库
SIFT
最小二乘法
image registration
automatic registration
GCP database
SIFT
least square method
