RPM-Net: Robust Point Matching using Learned Features

Abstract

Iterative Closest Point (ICP) solves the rigid point cloud registration problem iteratively in two steps: (1) make hard assignments of spatially closest point correspondences, and then (2) find the least-squares rigid transformation. The hard assignments of closest point correspondences based on spatial distances are sensitive to the initial rigid transformation and noisy/outlier points, which often cause ICP to converge to wrong local minima. In this paper, we propose the RPM-Net -- a less sensitive to initialization and more robust deep learning-based approach for rigid point cloud registration. To this end, our network uses the differentiable Sinkhorn layer and annealing to get soft assignments of point correspondences from hybrid features learned from both spatial coordinates and local geometry. To further improve registration performance, we introduce a secondary network to predict optimal annealing parameters. Unlike some existing methods, our RPM-Net handles missing correspondences and point clouds with partial visibility. Experimental results show that our RPM-Net achieves state-of-the-art performance compared to existing non-deep learning and recent deep learning methods. Our source code is available at the project website https://github.com/yewzijian/RPMNet .

10 pages, 4 figures. To appear in CVPR2020

References in corpus (9)

• Adam: A Method for Stochastic Optimization
• Open3D: A Modern Library for 3D Data Processing
• Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration
• 3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration
• DeepICP: An End-to-End Deep Neural Network for 3D Point Cloud Registration
• Dynamic Graph CNN for Learning on Point Clouds
• PRNet: Self-Supervised Learning for Partial-to-Partial Registration
• 3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions
• PCRNet: Point Cloud Registration Network using PointNet Encoding

Cited by in corpus (44)

• A comprehensive survey on point cloud registration
• CorrI2P: Deep Image-to-Point Cloud Registration via Dense Correspondence
• KSS-ICP: Point Cloud Registration based on Kendall Shape Space
• VRNet: Learning the Rectified Virtual Corresponding Points for 3D Point Cloud Registration
• A New Outlier Removal Strategy Based on Reliability of Correspondence Graph for Fast Point Cloud Registration
• Robust Point Cloud Registration Framework Based on Deep Graph Matching(TPAMI Version)
• CaSPR: Learning Canonical Spatiotemporal Point Cloud Representations
• Deep Shells: Unsupervised Shape Correspondence with Optimal Transport
• Deep Learning for 3D Point Cloud Understanding: A Survey
• Multi-Features Guidance Network for partial-to-partial point cloud registration
• Point Cloud Registration using Representative Overlapping Points
• PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency
• Category-Independent Articulated Object Tracking with Factor Graphs
• SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration
• Robust Point Cloud Registration Framework Based on Deep Graph Matching
• Robust Partial-to-Partial Point Cloud Registration in a Full Range
• Correspondence-Free Point Cloud Registration with SO(3)-Equivariant Implicit Shape Representations
• Correspondence Matrices are Underrated
• DBDNet:Partial-to-Partial Point Cloud Registration with Dual Branches Decoupling
• PREDATOR: Registration of 3D Point Clouds with Low Overlap
• Event-based Camera Simulation using Monte Carlo Path Tracing with Adaptive Denoising
• Searching Dense Point Correspondences via Permutation Matrix Learning
• Multiview Point Cloud Registration via Optimization in an Autoencoder Latent Space
• DeepI2P: Image-to-Point Cloud Registration via Deep Classification
• PointRegGPT: Boosting 3D Point Cloud Registration using Generative Point-Cloud Pairs for Training
• Sampling Network Guided Cross-Entropy Method for Unsupervised Point Cloud Registration
• GenReg: Deep Generative Method for Fast Point Cloud Registration
• Leveraging Cycle-Consistent Anchor Points for Self-Supervised RGB-D Registration
• Planning with Learned Dynamic Model for Unsupervised Point Cloud Registration
• A Helping (Human) Hand in Kinematic Structure Estimation
• A Direct Semi-Exhaustive Search Method for Robust, Partial-to-Full Point Cloud Registration
• Learning Scene Dynamics from Point Cloud Sequences
• CorrNet3D: Unsupervised End-to-end Learning of Dense Correspondence for 3D Point Clouds
• Cascading Feature Extraction for Fast Point Cloud Registration
• A Method to Generate High Precision Mesh Model and RGB-D Datasetfor 6D Pose Estimation Task
• PRANet: Point Cloud Registration with an Artificial Agent
• Attention-guided reference point shifting for Gaussian-mixture-based partial point set registration
• 3D Unsupervised Region-Aware Registration Transformer
• A Robust Loss for Point Cloud Registration
• Sinkhorn-CPD: Robust point cloud registration via unbalanced entropic optimal transport
• Correspondence-Free, Function-Based Sim-to-Real Learning for Deformable Surface Control
• Consistent Two-Flow Network for Tele-Registration of Point Clouds
• 3D Registration for Self-Occluded Objects in Context
• What Stops Learning-based 3D Registration from Working in the Real World?