Robust surface reconstruction from highly noisy point clouds using distributed elastic networks
- PDF / 1,665,282 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 40 Downloads / 172 Views
(0123456789().,-volV)(0123456789(). ,- volV)
EXTREME LEARNING MACHINE AND DEEP LEARNING NETWORKS
Robust surface reconstruction from highly noisy point clouds using distributed elastic networks Zhenghua Zhou1 Received: 25 January 2019 / Accepted: 30 July 2019 Ó Springer-Verlag London Ltd., part of Springer Nature 2019
Abstract In this paper, a novel distributed elastic random weights network (DERWN) is proposed to achieve robust surface reconstruction from highly noisy point clouds sampled from real surface. The designed elastic regularization with l1 and l2 penalty items makes the network more resilient to noise and effectively capture the intrinsic shape of surface. Sparsity constraints of output weight vectors and threshold-based nodes removal are conducive to determining appropriate number of hidden nodes of network and optimizing the distribution of hidden nodes. The distributed optimization manner in DERWN on the basis of alternating direction method of multipliers solves the problem that traditional RWN learning algorithm suffers from the limitation of memory with large-scale data. The proposed DERWN achieves a solution to global problem by solving local subproblems coordinately. Experimental results show that the proposed DERWN algorithm can robustly reconstruct the unknown surface in case of highly noisy data with satisfying accuracy and smoothness. Keywords Surface reconstruction Random weights network Elastic regularization Sparsity Distributed ADMM
1 Introduction Surface reconstruction from sample point clouds has been widely applied in computer-aided design (CAD) design, virtual reality, medical imaging, etc. Its goal is to create a surface model approximating the real model. The desired surface reconstruction algorithm is required to be capable of recovering both topology and geometry to fit the data correctly. However, these data sampled from the real surface always contain high noises induced by measurement error, and even some outliers which are far away from the real surface [1, 2]. When the above problems become serious, most existing algorithms cannot effectively and efficiently establish an accurate surface model only by utilizing the local areas of a point clouds independently. Furthermore, these methods usually impose strong assumptions on the original surface and its sample points. For this case, it is indispensable to take a global view in reconstruction algorithms. The goal of this paper is to & Zhenghua Zhou [email protected] 1
develop a robust surface reconstruction algorithm that can effectively handle noisy data sets from a global perspective. A novel algorithm, named distributed elastic random weights networks (DERWN) method, is proposed to extract useful information from highly noisy data for surface reconstruction. Unlike conventional surface reconstruction approaches that require specific information such as surface normals or pose demand on data accuracy and density, the proposed DERWN takes a set of noise sample points as input of neural network without any normal infor
Data Loading...
