Application of Improved SFM Adaptive Threshold Algorithm in Automatic 3D Reconstruction of Remote Sensing Images

Abstract

Motion recovery structure method has become a research hotspot of 3D reconstruction because of its advantages of low equipment cost, easy to carry, wide range of applicable scenes and so on. This paper proposes to use RANSAC adaptive threshold method to improve the SFM model estimation process, and proposes to combine the feature-based matching (FBM) method and the area based matching (ABM) method with NCC (normalized cross correlation) and LSM (least squares image matching) to perform FBM at the top of the image pyramid, Then, the location of matching points is adjusted layer by layer by using the region matching method with improved NCC and LSM, and finally a high-precision matching location is obtained. The modified matching points are used for basic matrix f estimation, so that the threshold can be adaptive to various scenes and data sets, and improve the accuracy of model estimation and final three dimensional reconstructions.