Fully Automatic Registration of 3D Point Clouds

By Ameesh Makadia, Alexander Patterson, Kostas Daniilidis. Computer Vision and Pattern Recognition, 2006

365 citations recorded by the source corpus.

We propose a novel technique for the registration of 3D point clouds which makes very few assumptions: we avoid any manual rough alignment or the use of landmarks, displacement can be arbitrarily large, and the two point sets can have very little overlap. Crude alignment is achieved by estimation of the 3D-rotation from two Extended Gaussian Images even when the data sets inducing them have partial overlap. The technique is based on the correlation of the two EGIs in the Fourier domain and makes use of the spherical and rotational harmonic transforms. For pairs with low overlap which fail a critical verification step, the rotational alignment can be obtained by the alignment of constellation images generated from the EGIs. Rotationally aligned sets are matched by correlation using the Fourier transform of volumetric functions. A fine alignment is acquired in the final step by running Iterative Closest Points with just few iterations.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Geology

View Paper: https://www.semanticscholar.org/paper/ad327e67c11eea75a6117cd8ed8871acc8b4bc39

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

In Geology

Citations in the corpus

Heritage Recording and 3D Modeling with Photogrammetry and 3D Scanning (2011, methodology): “The automated pair-wise coarse registration is normally performed using feature-based approaches [114-117] but in practical projects involving large and complex datasets, the coarse registration is still a manually driven procedure.”
Aligning point cloud views using persistent feature histograms (2008, methodology): “Similar systems have already been proposed such as in [22], where a rough initial guess is computed using Extended Gaussian Images (EGI) and the rotational Fourier transform.”
Deep Global Registration (2020, methodology): “[26] propose an iterative procedure to minimize correlation scores.”
Towards 3D Point cloud based object maps for household environments (2008, methodology): “Furthermore, methods for generating initial approximate alignments [10,35,19], as well as choosing alternate optimization methods [14,6], improved the results of the registration process.”
Data-driven grasping with partial sensor data (2009, methodology): “[21] correlated Extended Gaussian Images of two point clouds using spherical harmonics, and achieved excellent results but recommended a minimum of 45% overlap between”
Go-ICP: Solving 3D Registration Efficiently and Globally Optimally (2013, methodology): “Methods that make use of local invariant shape descriptors (e.g. spin image [19], shape contexts [4], EGI [24]) are mostly heuristic and do not address the optimality issue .”
Fast Global Registration (2016, methodology): “The typical workflow consists of two stages: global alignment, which computes an initial estimate of the rigid motion between two surfaces, followed by local refinement, which refines this initial estimate to obtain a tight registration [7, 12, 19, 25, 27, 37, 39].”
Non-Rigid Point Set Registration by Preserving Global and Local Structures (2015, background): “With a small number of transformation parameters, rigid registration is relatively easy and has been widely studied [1], [2], [8]–[10].”
Fast Point Feature Histograms (FPFH) for 3D registration (2009, methodology): “The ICP method has seen many improvements from its original form, from using non-linear optimization methods [7], [8], finding good initial guesses [9], [10], or es-”
Towards 3D lidar point cloud registration improvement using optimal neighborhood knowledge (2013, methodology): “If not, when scan orientations are unknown (e.g., terrestrial surveys), methods specific to the areas of interest are required (Makadia et al., 2006; Barnea and Filin, 2008; Theiler and Schindler, 2012).”
PointNetLK: Robust & Efficient Point Cloud Registration Using PointNet (2019, background): “For instance, scale invariant curvature descriptors [9], oriented descriptors [10], extended Gaussian images [19], fast point feature histograms [32], color intensity-based descriptors [11], global point signatures [6], heat kernels [25], etc.”
Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration (2016, methodology): “However, the joint problem cannot be trivially solved.”