Real-Time Large-Scale Dense 3D Reconstruction with Loop Closure
In the highly active research field of dense 3D reconstruction and modelling, loop closure is still a largely unsolved problem. While a number of previous works show how to accumulate keyframes, globally optimize their pose on closure, and compute a dense
- PDF / 7,652,487 Bytes
- 17 Pages / 439.37 x 666.142 pts Page_size
- 46 Downloads / 224 Views
Abstract. In the highly active research field of dense 3D reconstruction and modelling, loop closure is still a largely unsolved problem. While a number of previous works show how to accumulate keyframes, globally optimize their pose on closure, and compute a dense 3D model as a post-processing step, in this paper we propose an online framework which delivers a consistent 3D model to the user in real time. This is achieved by splitting the scene into submaps, and adjusting the poses of the submaps as and when required. We present a novel technique for accumulating relative pose constraints between the submaps at very little computational cost, and demonstrate how to maintain a lightweight, scalable global optimization of submap poses. In contrast to previous works, the number of submaps grows with the observed 3D scene surface, rather than with time. In addition to loop closure, the paper incorporates relocalization and provides a novel way of assessing tracking quality.
1
Introduction
The prompt delivery of highly detailed dense reconstructions of the 3D environment is currently one of the major frontiers in computer vision [1–6]. While off-line dense reconstruction and real-time sparse reconstruction are quite mature (e.g. [7–10]) it is only recently that real-time dense modelling and reconstruction have become feasible. There are two obvious enablers and catalysts. The first is the release of affordable RGB-D sensors such as the Kinect, and the second is the opening up of graphics processing units for general purpose computing. As the area develops, one can discern the research focus shifting from basic processing to obtain any reconstruction to more refined aspects of 3D representation and modelling. Earlier works such as [2,11,12] have firmly established volumetric representations as a powerful approach. Their memory requirements have been reduced by an order of magnitude more recently [13–17], rendering them attractive for large-scale mapping. While the competing surfel based representations (e.g. [18]) never had this issue, two common problems arising in all large scale mapping tasks are those of tracking drift and loop closure. Electronic supplementary material The online version of this chapter (doi:10. 1007/978-3-319-46484-8 30) contains supplementary material, which is available to authorized users. c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part VIII, LNCS 9912, pp. 500–516, 2016. DOI: 10.1007/978-3-319-46484-8 30
Real-Time Large-Scale Dense 3D Reconstruction
501
Fig. 1. Example reconstruction. The last row shows the results before and after loop closure, with images 1–10 showing constituent submaps.
Among the current methods addressing loop closure for dense 3D reconstruction, a first group [19–22] selects a subset of keyframes and gathers constraints between them to solve the alignment problem, and only then considers dense reconstruction as a post-processing step. For this group, dense reconstruction typically has to be repeated from scratch whenever there are ch
Data Loading...
