A Registration Method for 3D Point Clouds with Convolutional Neural Network

Viewpoint independent 3D object pose estimation is one of the most fundamental step of position based vision servo, autopilot, medical scans process, reverse engineering and many other fields. In this paper, we presents a new method to estimate 3D pose using the convolutional neural network (CNN), which can apply to the 3D point cloud arrays. An interest point detector was proposed and interest points were computed in both source and target point clouds by region growing cluster method during offline training of CNN. Rather than matching the correspondences by rejecting and filtering iteratively, a CNN classification model is designed to match a certain subset of correspondences. And a 3D shape representation of interest points was projected onto an input feature map which is amenable to CNN. After aligning point clouds according to the prediction made by CNN, iterative closest point (ICP) algorithm is used for fine alignment. Finally, experiments were conducted to show the proposed method was effective and robust to noise and point cloud partial missing.

[1]  Darius Burschka,et al.  An Efficient RANSAC for 3D Object Recognition in Noisy and Occluded Scenes , 2010, ACCV.

[2]  François Chaumette,et al.  Visual servo control. I. Basic approaches , 2006, IEEE Robotics & Automation Magazine.

[3]  Mohammed Bennamoun,et al.  On the Repeatability and Quality of Keypoints for Local Feature-based 3D Object Retrieval from Cluttered Scenes , 2009, International Journal of Computer Vision.

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

[5]  Alexandre Boulch,et al.  Deep Learning for Robust Normal Estimation in Unstructured Point Clouds , 2016, Comput. Graph. Forum.

[6]  Tianshi Chen,et al.  A new algorithm for non-rigid point matching using geodesic graph model , 2015, 2015 IEEE International Conference on Mechatronics and Automation (ICMA).

[7]  Yu Zhong,et al.  Intrinsic shape signatures: A shape descriptor for 3D object recognition , 2009, 2009 IEEE 12th International Conference on Computer Vision Workshops, ICCV Workshops.

[8]  Jun Cheng,et al.  Registration for 3-D point cloud using angular-invariant feature , 2009, Neurocomputing.

[9]  Hui Chen,et al.  3D free-form object recognition in range images using local surface patches , 2004, Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004..

[10]  Nassir Navab,et al.  Model globally, match locally: Efficient and robust 3D object recognition , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

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

[12]  Joaquim Salvi,et al.  A Qualitative Review on 3D Coarse Registration Methods , 2015, ACM Comput. Surv..

[13]  Z. Jane Wang,et al.  A CNN Regression Approach for Real-Time 2D/3D Registration , 2016, IEEE Transactions on Medical Imaging.

[14]  K. S. Arun,et al.  Least-Squares Fitting of Two 3-D Point Sets , 1987, IEEE Transactions on Pattern Analysis and Machine Intelligence.

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

[16]  Federico Tombari,et al.  Performance Evaluation of 3D Keypoint Detectors , 2011, 2011 International Conference on 3D Imaging, Modeling, Processing, Visualization and Transmission.