A tiny facet primitive remote sensing image registration algorithm based on optimized Delaunay Triangulation

To solve the accuracy problem of remote sensing image registration which is influenced by the uneven distribution of control points, a tiny facet primitive remote sensing image registration algorithm based on the optimized Delaunay Triangulation is proposed in this paper. Firstly, the Rational Function Model(RFM) and Digital Surface Model(DSM) are used for image ortho rectification. Secondly, feature points are matched through the SIFT operator, and optimized by the RANSAC algorithm. At the same time, the edge grid points are extracted, and the affine transformation, epipolar constraint and gray similarity constraint are used to match them. Then the initial Delaunay Triangulation is built, and based on the triangle area constraint and triangle angle optimized that. Finally, the tiny facet primitive image registration is achieved through the scan line filling algorithm. In this paper, simultaneous and multi-temporal remote sensing images are taken as data source to conduct the experiment. The experimental results of sets of data show that the applicability and validity, and accuracy of image registration can reach sub-pixel level, and effectively solves the problem which exists in remote sensing images with complicated geometric deformation.