三角网约束的立体影像密集匹配方法  

Stereo Image Dense Matching Method With Triangle Net Constraint

作　　者：王琦[1] 宋伟东[1] 王竞雪[1] WANG Qi;SONG Weidong;WANG Jingxue(School of Geomatics, Liaoning Technical University, Fuxin 123000, China)

机构地区：[1]辽宁工程技术大学测绘与地理与地理科学学院,辽宁阜新123000

出　　处：《测绘与空间地理信息》2019年第5期237-240,共4页Geomatics & Spatial Information Technology

摘　　要：文中总结了基于三角网约束的立体影像密集匹配方法。本次实验采用核线影像,首先在核线影像上使用SIFT算子匹配出一部分可靠的种子点,然后用这些种子点构建Delaunay三角网。该网络能够将所加密的点的位置范围作为一个约束,根据左右影像上同名三角网的位置、左影像上某一三角形重心位置、该三角形在右影像上的对应的三角形的重新坐标,以及左侧重心在右侧像方影像的核线,在核线约束与灰度相似性约束的两层限制下,完成点-点的匹配工作。本文采用Matlab程序实现对待匹配三角形重心的加密,运用核线约束方法,将点匹配的搜索区域由二维降到一维,不但能缩小搜索范围,提高工作效率,而且很大程度上降低了错误发生的概率,既能使得算法的运算速度得以提升,又能使其匹配精度得到较大的改善。A stereo image dense matching method based on triangular net constraint is proposed considering the sparse problem of the same name with the SIFT algorithm, and epipo images are used in this experiment. Firstly, the SIFT operator is used to match a number of reliable seed points on the epipo images, which will be constructed to the Delaunay triangulations. According to the same name triangles on both left and right images, the gravity centers of them and the information about epipo lines, point-point matching should be processed in the triangulations constrained area with the conditions of both epipo line constraint and gray similarity constraint. In this paper, Matlab program is adopted to realize the coarctation of the center of the matching triangle. Using the epipo line constraint, the searching area of the point matching is reduced to one dimension from two. It can not only narrow the range of searching, improve the work efficiency, but also reduce the probability of errors, accelerate the calculation speed and improve the matching precision.

关 键 词：密集匹配 DELAUNAY三角网 重心 核线约束 

分 类 号：P209[天文地球—测绘科学与技术]