RepMatch: Robust Feature Matching and Pose for Reconstructing Modern Cities
A perennial problem in recovering 3-D models from images is repeated structures common in modern cities. The problem can be traced to the feature matcher which needs to match less distinctive features (permitting wide-baselines and avoiding broken sequenc
- PDF / 4,144,485 Bytes
- 18 Pages / 439.37 x 666.142 pts Page_size
- 6 Downloads / 207 Views
3
Advanced Digital Sciences Center, Singapore, Singapore [email protected], [email protected] 2 Institute of Infocomm Research, Singapore, Singapore University of Illinois Urbana-Champagne, Champaign, USA 4 Simon Fraser University, Burnaby, Canada
Abstract. A perennial problem in recovering 3-D models from images is repeated structures common in modern cities. The problem can be traced to the feature matcher which needs to match less distinctive features (permitting wide-baselines and avoiding broken sequences), while simultaneously avoiding incorrect matching of ambiguous repeated features. To meet this need, we develop RepMatch, an epipolar guided (assumes predominately camera motion) feature matcher that accommodates both wide-baselines and repeated structures. RepMatch is based on using RANSAC to guide the training of match consistency curves for differentiating true and false matches. By considering the set of all nearest-neighbor matches, RepMatch can procure very large numbers of matches over wide baselines. This in turn lends stability to pose estimation. RepMatch’s performance compares favorably on standard datasets and enables more complete reconstructions of modern architectures.
Keywords: Structure from motion
1
· Correspondence · RANSAC
Introduction
Structure-from-Motion or SfM is the recovery of 3-D structure from image sets. Over the years, SfM has made remarkable progress. Current technology can create impressively large scale reconstructions, a signature achievement being the reconstruction of ancient Rome by leveraging the abundance of Internet images [1]. However, SfM systems have difficulty reconstructing modern buildings from small, user-captured datasets. J. Lu—This study is supported by the HCCS grant at ADSC from Singapore’s Agency for Science, Technology and Research. Electronic supplementary material The online version of this chapter (doi:10. 1007/978-3-319-46448-0 34) contains supplementary material, which is available to authorized users. c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part I, LNCS 9905, pp. 562–579, 2016. DOI: 10.1007/978-3-319-46448-0 34
RepMatch
563
Input Images
(a) Visual SfM
(b) Visual SfM with our matches
(c) Dense reconstruction
Fig. 1. 3-D reconstruction on modern buildings. (a) Screen shot of Visual SfM [2], a classic 3-D reconstruction system which splinters the sequence into 4 segments. (b) The same reconstruction system with our matches forms a complete loop. (c) The pose estimated in (b) is sufficiently accurate for high quality dense reconstruction.
The problem stems from SfM’s dependence on feature matching from which camera position (pose) and 3-D structure are inferred. Feature matching needs to procure large numbers of wide-baseline matches to prevent image sequences from splintering. Yet, it must also be robust to repetitive structures. Unfortunately, modern urban environments contain both challenges in abundance. Trees and other occluders limit available view-points, necessitating matching wid
Data Loading...
