3D Local Features for Direct Pairwise Registration

Abstract

We present a novel, data driven approach for solving the problem of registration of two point cloud scans. Our approach is direct in the sense that a single pair of corresponding local patches already provides the necessary transformation cue for the global registration. To achieve that, we first endow the state of the art PPF-FoldNet [19] auto-encoder (AE) with a pose-variant sibling, where the discrepancy between the two leads to pose-specific descriptors. Based upon this, we introduce RelativeNet, a relative pose estimation network to assign correspondence-specific orientations to the keypoints, eliminating any local reference frame computations. Finally, we devise a simple yet effective hypothesize-and-verify algorithm to quickly use the predictions and align two point sets. Our extensive quantitative and qualitative experiments suggests that our approach outperforms the state of the art in challenging real datasets of pairwise registration and that augmenting the keypoints with local pose information leads to better generalization and a dramatic speed-up.

Learning Multiview 3D Point Cloud Registration

“…that our approach can achieve ≈ 4 percentage points higher recall than state-of-the-art without being trained on synthetic data and thus confirming the good generalization capacity of our approach. Note that while the average precision across the scenes is low for all the methods, several works [14,38,22] show that the precision can easily be increased using pruning without almost any loss in the recall.…”

“…We begin by evaluating the pairwise registration part of our algorithm on a traditional geometric registration task. We compare the results of our method to the state-of-the-art data-driven feature descriptors 3DMatch [67], CGF [38], PPFNet [21], 3DSmoothNet (3DS) [28], and FCGF [16], which is also used as part of our algorithm, as well as to a recent network based registration algorithm 3DR [22]. Following the evaluation procedure of 3DMatch [67] we complement all the descriptor based methods with the RANSAC-based transformation parameter estimation.…”

“…We follow the evaluation protocol of [14] and report the average registration recall and precision across all four scenes. We compare our approach to the recent data driven approaches 3DMatch [67], CGF [38]+FGR [69] or CZK [14], RelativeNet (RN) [22], 3DR [22] and traditional methods CZK [14] and Latent RANSAC (LR) [40]. Fig.…”

End-to-End CAD Model Retrieval and 9DoF Alignment in 3D Scans

“…Traditionally, these descriptors were hand-crafted, and often based on a computation of histograms (e.g., point normals), such as FPFH [29], SHOT [35], or point-pair features [11]. More recently, with advances in deep neural networks, these descriptors can be learned, for instance based on an implicit signed distance field representation [40,9,10]. A typical pipeline for CAD-to-scan alignments builds on these descriptors; i.e., the first step is to find 3D feature matches and then use a variant of RANSAC or PnP to compute 6DoF or 9Dof CAD model alignments.…”

LRF-Net: Learning Local Reference Frames for 3D Local Shape Description and Matching

“…Traditional 6-DoF pose estimation is usually performed via RANSAC [ 6 ], which randomly selects inlier correspondences from an initial correspondence pool for pose prediction. Such random sampling method is neither reliable nor computational efficient [ 7 ]. By contrast, we can directly predict an initial pose via two corresponding LRFs, reducing the computational complexity from to .…”