Geometrically stable sampling for the ICP algorithm

By Natasha Gelfand, Szymon Rusinkiewicz, Leslie Ikemoto, Marc Levoy. Fourth International Conference on 3-D Digital Imaging and Modeling, 2003. 3DIM 2003. Proceedings., 2003

422 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/eed3b80ebdb4868f62cf7fcfb35bd713067a7325

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 Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

Citations in the corpus

RPM-Net: Robust Point Matching Using Learned Features (2020, background): “, selecting suitable points [10, 25] or weighting point correspondences [11].”
Colored Point Cloud Registration Revisited (2017, background): “A common failure mode of ICP is instability in the presence of smooth surfaces [14, 46].”
Blending liquids (2014, methodology): “The iterated closest point method (ICP) was introduced by Besl and McKay [1992] and numerous improvements have been suggested over the years [Hähnel et al. 2003; Gelfand et al. 2003; Rusinkiewicz and Levoy 2001; Brown and Rusinkiewicz 2007].”
On the Repeatability and Quality of Keypoints for Local Feature-based 3D Object Retrieval from Cluttered Scenes (2009, methodology): “In the case of [10], there may be ambiguity in 3 rotations and 3 translations and hence a 6x6 covariance matrix is used.”
Scan registration for autonomous mining vehicles using 3D‐NDT (2007, methodology): “One way to speed up the search is to use an efficient search data structure, such as a kd-tree with approximate nearest-neighbor search ͑Greenspan & Yurick, 2003͒, but the search pass is still the main bottleneck for the algorithm's running time.”
Fast Point Feature Histograms (FPFH) for 3D registration (2009, background): “In large datasets, the number of points with the same or similar PFH might be very big and could lead to the so called “sliding” problem in registration, where points on certain surfaces do not contribute positively to the global distance metric due to ambiguous correspondences [17].”
A robust RGB-D SLAM algorithm (2012, background): “This can be associated with a phenomenon known as sliding [17], where ICP cannot determine where the best fit lies along a plane.”
IMLS-SLAM: Scan-to-Model Matching Based on 3D Data (2018, methodology): “[10] gives a more interesting strategy using the covariance matrix of the scan point cloud to find geometric stable sampling.”
RigidFusion: RGB‐D Scene Reconstruction with Rigidly‐moving Objects (2021, background): “However, the motion cues can be ambiguous, especially when the camera are moving, and the local patches of indoor objects are usually under-constrained [GIRL03].”
Registration of 3D Point Clouds and Meshes: A Survey from Rigid to Nonrigid (2013, neutral): “D et er m in is tic Gradient Descent & related methods [11], [50]–[52], [54], [62], [93] [24], [28], [30], [34] Newton, Gauss-Newton, L-M method [14], [46], [66], [94] [2], [31], [32], [79], [80]”
Matching 2.5D face scans to 3D models (2006, background): “The point-to-plane distance used in [34] makes the ICP algorithm less susceptible to local minima than the point-to-point metric [38].”