SIFT与粗差剔除算法相结合的机载SAR影像匹配研究
收稿日期: 2012-10-15
修回日期: 2013-01-04
网络出版日期: 2013-06-17
基金资助
国家“863”计划课题(2011AA120401);国家自然科学基金项目(41071237)。
Matching of Airborne SAR Images Based on a Combination of SIFT Algorithm with Mismatching Points Eliminated Algorithm
Received date: 2012-10-15
Revised date: 2013-01-04
Online published: 2013-06-17
SAR影像匹配是SAR数据处理的重要环节,但是,SAR影像匹配成功率、正确率及精度较低。通过SAR影像匹配,建立SAR匹配像对,对雷达立体观察和立体测量有重要的意义。在机载SAR影像匹配中,应用SIFT 算法,获得较稳定的特征,并结合2D单应变换的RANSAC算法剔除误匹配点对;选取三组不同类别的机载SAR影像数据,利用SIFT 和粗差剔除相结合的算法,通过VC++和OpenCV编程,提取出特征稳定且均匀分布的同名点对,完成机载SAR影像匹配实验。结果表明,对于含有人工建筑物的机载SAR影像,SIFT 算法可有效地提取大量稳定的匹配点对,其正确率高;对于含有自然植被的机载SAR影像,SIFT 算法可有效地提取较多稳定的匹配点对,其正确率较高;由于SIFT 算法实质上是基于局部灰度匹配的算法,对于纹理信息缺乏的机载SAR影像,只可提取少量稳定的匹配点对,其正确率相对前两者较低。总体而言,在机载SAR影像中SIFT 算法能够提取到稳定的匹配点对,并结合基于2D单应变换的RANSAC算法,可有效剔除误匹配点对,提高匹配正确率及精度。
关键词: 2D单应变换的RANSAC算法; 匹配; 机载SAR影像; SIFT算法
姜文聪, 张继贤, 程春泉, 梁勇 . SIFT与粗差剔除算法相结合的机载SAR影像匹配研究[J]. 地球信息科学学报, 2013 , 15(3) : 440 -444 . DOI: 10.3724/SP.J.1047.2013.00440
Success rate, correct rate and precision of SAR image matching are usually low, so it indirectly affects the subsequent application of SAR images. SIFT feature is the local feature of the image, and it remains invariance of the rotation, scale and luminance variation; it also maintains a certain degree of stability of the viewing angle variation, the affine transformation and the noise. The stable features are extracted by using SIFT algorithm which is applied to airborne SAR image matching. AS 2D homography transform is the model of the RANSAC algorithm, four pairs of points at least are used to calculate 2D homography transform model parameters. RANSAC algorithm is a kind of robust parameter estimation method to obtain the effective sample data. The mismatching points are eliminated by using the RANSAC algorithm based on 2D homograpy transform. Based on VC++ and OpenCV programming, three sets of different kinds of airborne SAR image data are selected to complete matching experiment of airborne SAR image in which SIFT algorithm combined with error eliminated algorithm is used. The result shows that for airborne SAR images of artificial buildings and natural vegetation, SIFT algorithm can be used to effectively extract numerous stable matching points and the correct rate is high; for airborne SAR images of lack of texture information, SIFT algorithm can be used to extract a small amount of stable matching points and the correct rate is obviously lower than the former. In a word, SIFT algorithm can be used to extract the stable matching points in the airborne SAR image matching, and the mismatching points are effectively eliminated by using RANSAC algorithm based on 2D homograpy transform, so the correct rate and precision of matching is greatly improved.
[1] 尤红建,付琨.合成孔径雷达图像精准处理[M].北京:科学出版社,2011.
[2] 吕文涛,吕高焕.SIFT 算法在雷达图像匹配中的应用[J].信息与电子工程,2010,8(4):388-392.
[3] Lowe D G. Distinctive image features from Scale-invariantkeypoints[J]. International Journal of Computer Vision,2004,60(2):91-110.
[4] Lowe D G. Object recognition from local scale-invariant features[C]. International Conference on Computer Vision,Washington, USA, IEEE Computer Society, 1999.
[5] Lowe D. Local feature view clustering for 3D object recognition[EB/OL].
[6] 曹红杏.基于特征的图像拼接技术研究[D]. 北京:中国科学院研究生院,2008.
[7] 刘景正.基于特征的SAR影像匹配技术研究[D].郑州:解放军信息工程大学,2007.
[8] 于秋则.合成孔径雷达图像匹配导航技术研究[D].武汉:华中科技大学,2004.
[9] Li Q L,Wang G Y,Liu J G,et al. Robust scale-invariantfeature matching for remote sensing image registration[J].IEEE Geoscience and Remote Sensing Letters(S1545-598X), 2009, 6(2): 287-291.
[10] 王琳.高精度、高可靠的无人机影像全自动相对定向及模型连接研究[D].北京:中国测绘科学研究院,2008.
[11] 罗晓曼.合成孔径雷达影像匹配技术研究[D].北京:中国测绘科学研究院,2008.
[12] 陈尔学,李增元,田昕,等.尺度不变特征变换法在SAR影像匹配中的应用[J].自动化学报,2008,34(8):861-868.
[13] 宋巨艳.RANSAC 算法及其在遥感图像处理中的应用[D].北京:华北电力大学,2011.
[14] 孙艳丽,等.SIFT 在高分辨率SAR图像自动配准中的性能分析[J].电子设计工程,2011,19(7):180-183.
[15] 杨朝辉,陈鹰,邵永社,等.基于SIFT 特征的合成孔径雷达景象匹配方法[J].计算机应用,2008,28(9):2404-2406.
/
〈 | 〉 |