CorsNet: 3D Point Cloud Registration by Deep Neural Network

Point cloud registration is a key problem for robotics and computer vision communities. This represents estimating a rigid transform which aligns one point cloud to another. Iterative closest point (ICP) is a well-known classical method for this problem, yet it generally achieves high alignment only when the source and template point cloud are mostly pre-aligned. If each point cloud is far away or contains a repeating structure, the registration often fails because of being fallen into a local minimum. Recently, inspired by PointNet, several deep learning-based methods have been developed. PointNetLK is a representative approach, which directly optimizes the distance of aggregated features using gradient method by Jacobian. In this paper, we propose a point cloud registration system based on deep learning: CorsNet. Since CorsNet concatenates the local features with the global features and regresses correspondences between point clouds, not directly pose or aggregated features, more useful information is integrated than the conventional approaches. For comparison, we also developed a novel deep learning approach (DirectNet) that directly regresses the pose between point clouds. Through our experiments, we show that CorsNet achieves higher accuracy than not only the classic ICP method, but also the recently proposed learning-based proposal PointNetLK and DirectNet, including on seen and unseen categories.

[1]  Andrew E. Johnson,et al.  Using Spin Images for Efficient Object Recognition in Cluttered 3D Scenes , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[2]  Jianxiong Xiao,et al.  3D ShapeNets: A deep representation for volumetric shapes , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[3]  Heinz Hügli,et al.  A multi-resolution scheme ICP algorithm for fast shape registration , 2002, Proceedings. First International Symposium on 3D Data Processing Visualization and Transmission.

[4]  Yasuhiro Aoki,et al.  PointNetLK: Robust & Efficient Point Cloud Registration Using PointNet , 2019, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[5]  Leonidas J. Guibas,et al.  One Point Isometric Matching with the Heat Kernel , 2010, Comput. Graph. Forum.

[6]  Chin Seng Chua,et al.  Point Signatures: A New Representation for 3D Object Recognition , 1997, International Journal of Computer Vision.

[7]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[8]  Yaron Lipman,et al.  Point registration via efficient convex relaxation , 2016, ACM Trans. Graph..

[9]  Mohammed Bennamoun,et al.  3D Object Recognition in Cluttered Scenes with Local Surface Features: A Survey , 2014, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[10]  Daniel Cohen-Or,et al.  4-points congruent sets for robust pairwise surface registration , 2008, ACM Trans. Graph..

[11]  Jiaolong Yang,et al.  Go-ICP: Solving 3D Registration Efficiently and Globally Optimally , 2013, 2013 IEEE International Conference on Computer Vision.

[12]  Zi Jian Yew,et al.  3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration , 2018, ECCV.

[13]  Leonidas J. Guibas,et al.  PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[14]  Leonidas J. Guibas,et al.  PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space , 2017, NIPS.

[15]  Zhengyou Zhang,et al.  Microsoft Kinect Sensor and Its Effect , 2012, IEEE Multim..

[16]  Nico Blodow,et al.  Fast Point Feature Histograms (FPFH) for 3D registration , 2009, 2009 IEEE International Conference on Robotics and Automation.

[17]  Gary R. Bradski,et al.  Monte Carlo Pose Estimation with Quaternion Kernels and the Bingham Distribution , 2011, Robotics: Science and Systems.

[18]  Joel W. Burdick,et al.  Convex relaxations of SE(2) and SE(3) for visual pose estimation , 2014, 2014 IEEE International Conference on Robotics and Automation (ICRA).

[19]  Marc Levoy,et al.  Efficient variants of the ICP algorithm , 2001, Proceedings Third International Conference on 3-D Digital Imaging and Modeling.

[20]  Andrea Tagliasacchi,et al.  Eurographics Symposium on Geometry Processing 2013 Sparse Iterative Closest Point , 2022 .

[21]  Jiaolong Yang,et al.  Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[22]  James S. Duncan,et al.  A Robust Point Matching Algorithm for Autoradiograph Alignment , 1996, VBC.

[23]  Paul J. Besl,et al.  A Method for Registration of 3-D Shapes , 1992, IEEE Trans. Pattern Anal. Mach. Intell..

[24]  Russ Tedrake,et al.  Globally Optimal Object Pose Estimation in Point Clouds with Mixed-Integer Programming , 2017, ISRR.

[25]  Leonidas J. Guibas,et al.  Frustum PointNets for 3D Object Detection from RGB-D Data , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[26]  Takeo Kanade,et al.  An Iterative Image Registration Technique with an Application to Stereo Vision , 1981, IJCAI.

[27]  Leonidas J. Guibas,et al.  Robust global registration , 2005, SGP '05.

[28]  Howie Choset,et al.  Probabilistic pose estimation using a Bingham distribution-based linear filter , 2018, Int. J. Robotics Res..