Registration of 3D Point Clouds and Meshes: A Survey from Rigid to Nonrigid

Abstract

Abstract-3D surface registration transforms multiple 3D datasets into the same coordinate system so as to align overlapping components of these sets. Recent surveys have covered different aspects of either rigid or non-rigid registration, but seldom discuss them as a whole. Our study serves two purposes: (i) to give a comprehensive survey of both types of registration, focusing on 3D point clouds and meshes, and (ii) to provide a better understanding of registration from the perspective of data fitting. Registration is closely related to data fitting in that it comprises three core interwoven components: model selection, correspondences & constraints and optimization. Study of these components (i) provides a basis for comparison of the novelties of different techniques, (ii) reveals the similarity of rigid and non-rigid registration in terms of problem representations, and (iii) shows how over-fitting arises in non-rigid registration and the reasons for increasing interest in intrinsic techniques. We further summarise some practical issues of registration which include initializations and evaluations, and discuss some of our own observations, insights and foreseeable research trends.

Semi-automatic landmark point annotation for geometric morphometrics

“…A potentially simpler alternative when manually annotated training data is available, and the approach adopted here, is to consider the mappings between landmark points as sparse transformations from the coordinate systems of the database images to that of the query image, such that the problem falls into the general domain of registration. Registration, the estimation of a transformation that maps one image (the source) into the coordinate system of another image (the target) is a core problem in machine vision and medical image analysis with a correspondingly extensive literature; general reviews of medical image registration are provided by [13][14][15] and [16], whilst [17,18] and [19] provide recent reviews of surface registration algorithms, with particular reference to surfaces represented by point clouds or meshes.…”

Submap bathymetric SLAM using structured light in underwater environments

“…Also, further results in the evaluation of the SLAM solution will be presented, together with a parametrization study on the submap size, or whether smoothing out the submaps prior to registration, or using multiscale resolution as suggested in [20] is applicable. …”

Region-Based Correspondence Between 3D Shapes via Spatially Smooth Biclustering

“…The goal of most shape matching techniques is to find correspondences between points or regions in a given pair of shapes, which can differ by a variety of deformations. This includes both nearrigid correspondences, if the shapes are related by a rotation and a translation, and the more general and challenging non-rigid shape matching problem [42], in which shapes can undergo other transformations, such as articulated motion of humans. Although several methods have been proposed to address the latter problem, most techniques either use a prescribed deformation model (e.g., near-isometries or conformal deformations [3,17]), or rely on user-provided landmark correspondences [1].…”