摘要
针对航空影像中高程变化对视差影响较大,而传统的单应矩阵与几何约束未考虑视差影响的问题,该文提出了一种适合航空影像的直线匹配方法。首先采用尺度不变特征变换(SIFT)算法进行特征匹配,并计算立体影像间的基础矩阵,通过核线纠正消除其上下视差;然后计算影像间的单应矩阵,并对Hough变换提取的直线段进行变换,最终通过直线段间的端点距离作为约束条件实现了直线匹配。采用多组航空影像开展了实验,结果表明:针对航空影像中高程变化较大的区域,提出的方法可以减少视差的影响,匹配效果优于传统单应矩阵与几何约束的方法,且匹配可靠性高。
In view of the disparity influence caused by elevation change in aerial images,the traditional homography matrix and geometric constraint line matching method do not consider the disparity effect,a straight line matching method suitable for aerial images was proposed.Firstly,scale-invariant feature transform(SIFT)algorithm was used for feature extraction and matching.The basic matrix between stereo images was calculated by combining the matching results,and vertical disparity was eliminated by epipolar correction.Then the homography matrix between images was calculated and the straight line segment extracted by Hough transform was transformed.Finally,the straight line matching was realized through the endpoint distance between lines.Experiments were carried out with several aerial images.The results showed that,for regions with large elevation variations,the proposed method could reduce the parallax effect,and the matching effect was better than that of the traditional homography matrix and geometric constraints,and the matching reliability was high.
作者
徐辛超
李旭佳
马钰
刘少创
刘明岳
XU Xinchao;LI Xujia;MA Yu;LIU Shaochuang;LIU Mingyue(School of Gematics,Liaoning Technical University,Fuxin,Liaonng 123000,China;Institute of Remote Sensing and Digital Earth,Chinese Academy of Sciences,Beijing 100101,China)
出处
《测绘科学》
CSCD
北大核心
2020年第2期92-97,110,共7页
Science of Surveying and Mapping
基金
国家级大学生创新创业训练计划项目(201810147127)
国家自然科学基金项目(41401535)
国家重点研发计划项目(2016YFC0803102).
