Topology-Aware Non-Rigid Point Cloud Registration

Abstract

In this paper, we introduce a non-rigid registration pipeline for pairs of unorganized point clouds that may be topologically different. Standard warp field estimation algorithms, even under robust, discontinuity-preserving regularization, tend to produce erratic motion estimates on boundaries associated with 'close-to-open' topology changes. We overcome this limitation by exploiting backward motion: in the opposite motion direction, a 'close-to-open' event becomes 'open-to-close', which is by default handled correctly. At the core of our approach lies a general, topology-agnostic warp field estimation algorithm, similar to those employed in recently introduced dynamic reconstruction systems from RGB-D input. We improve motion estimation on boundaries associated with topology changes in an efficient post-processing phase. Based on both forward and (inverted) backward warp hypotheses, we explicitly detect regions of the deformed geometry that undergo topological changes by means of local deformation criteria and broadly classify them as 'contacts' or 'separations'. Subsequently, the two motion hypotheses are seamlessly blended on a local basis, according to the type and proximity of detected events. Our method achieves state-of-the-art motion estimation accuracy on the MPI Sintel dataset. Experiments on a custom dataset with topological event annotations demonstrate the effectiveness of our pipeline in estimating motion on event boundaries, as well as promising performance in explicit topological event detection. (1993) and an M.S. from the Graz University of Technology (1989), both in Applied Mathematics.Her research interest has been to understand principles of active vision systems and develop biological-inspired methods, especially in the area of motion. Her recent work has focused on human action interpretation and the development of event-based motion algorithms.

Topology-Change-Aware Volumetric Fusion for Dynamic Scene Reconstruction

“…Our approach can also successfully detect all topology change events and update the connectivity of EDG and TSDF grids to reflect such topology changes accordingly in reconstructed geometry. It is worth noting that our method can handle a more complex case like seq "alex (close to open)" (from [33]) -hand moving from contacting with body to no contact, which is not demonstrated in [45]. Besides that, Zampogiannis et al [45] did not address how to reconstruct the geometry of dynamic scenes under such topology changes, as will be introduced below.…”

“…It is worth noting that our method can handle a more complex case like seq "alex (close to open)" (from [33]) -hand moving from contacting with body to no contact, which is not demonstrated in [45]. Besides that, Zampogiannis et al [45] did not address how to reconstruct the geometry of dynamic scenes under such topology changes, as will be introduced below. [45].…”

“…Zampogiannis et al [45] proposed a topology-change-aware non-rigid point cloud registration approach by detecting topology change regions based on stretch and compression measurement in both forward and backward motion. However, how to recover the geometry of dynamic scenes under such topology changes is not explored.…”

“…In order to demonstrate that our topology change detection really works well, we run it on some public datasets used in [45], as shown in Fig. 2.…”

“…Besides that, Zampogiannis et al [45] did not address how to reconstruct the geometry of dynamic scenes under such topology changes, as will be introduced below. [45]. Row 4 is another challenging case from KillingFusion [33].…”