3d reconstruction using SFM with sequence images in hadoop

Abstract

Reconstructing a 3D structure from 2D images is a hot research topic in computer vision, and many approaches have been proposed to solve this problem in recent years. The structure from motion (SfM) algorithm has the potential to successfully reconstruct geometry from an image set, but the execution of the SfM algorithm on a single computer is only possible with small image sets. If an image set is large, the reconstruction speed decreases and it may even be too slow to complete 3D reconstruction. With the advent of cloud computing platforms, increasingly complex algorithms are being applied via cloud computing. Thus, the algorithm calculation speed for 3D reconstruction may be improved by taking advantage of the computational power of computer clusters. In this paper, we propose an algorithm for realizing 3D reconstruction from sequence images on the Hadoop cluster. Our test results reveal that, using Hadoop, we can achieve fast 3D reconstruction. © 2019, Success Culture Press. All rights reserved.