DIMENSIONALITY BASED SCALE SELECTION IN 3D LIDAR POINT CLOUDS

By J. Demantké, Clément Mallet, Nicolas David, B. Vallet. 2012

353 citations recorded by the source corpus.

Abstract. This papers presents a multi-scale method that computes robust geometric features on lidar point clouds in order to retrieve the optimal neighborhood size for each point. Three dimensionality features are calculated on spherical neighborhoods at various radius sizes. Based on combinations of the eigenvalues of the local structure tensor, they describe the shape of the neighborhood, indicating whether the local geometry is more linear (1D), planar (2D) or volumetric (3D). Two radius-selection criteria have been tested and compared for ﬁnding automatically the optimal neighborhood radius for each point. Besides, such procedure allows a dimensionality labelling, giving signiﬁcant hints for classiﬁcation and segmentation purposes. The method is successfully applied to 3D point clouds from airborne, terrestrial, and mobile mapping systems since no a priori knowledge on the distribution of the 3D points is required. Extracted dimensionality features and labellings are then favorably compared to those computed from constant size neighborhoods.

Subfields: Computer Graphics and Computer-Aided Design, Environmental Engineering, Geology, 3D Vision & Pose

How researchers see this paper

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

In Computer Graphics and Computer-Aided Design

In Environmental Engineering

In Geology

In 3D Vision & Pose

Citations in the corpus

PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space (2017, methodology): “Previously several approaches have been developed regarding this [19; 17; 2; 6; 7; 30] either in geometry processing community or photogrammetry and remote sensing community.”
Point cloud labeling using 3D Convolutional Neural Network (2016, methodology): “based on it have been applied in point-level classification tasks [6], [3].”
Towards 3D lidar point cloud registration improvement using optimal neighborhood knowledge (2013, methodology): “Since normal computation has been improved using the method of Demantké et al. (2011), this solution is also tested.”
Fast Automatic Precision Tree Models from Terrestrial Laser Scanner Data (2013, methodology): “The eigenvalues of S(B) can be used as indicators for the dimensionality of the set B [36], i.”
Semantic3D.net: A new Large-scale Point Cloud Classification Benchmark (2017, background): “Typical features encode surface properties based on the covariance tensor of a point’s neighborhood (Demantké et al., 2011) or a randomized set of histograms (Blomley et al.”
IMLS-SLAM: Scan-to-Model Matching Based on 3D Data (2018, background): “For every point we keep the planar scalar a2D of its neighborhood, as defined by [12]: a2D = (σ2 − σ3)/σ1 where σi = √ λi and λi are eigenvalues of the PCA for the normal computation (see [12] for more details).”
FAST SEMANTIC SEGMENTATION OF 3D POINT CLOUDS WITH STRONGLY VARYING DENSITY (2016, methodology): “Features designed for dense point cloud description are often based on the structure tensor of a point’s neighborhood (Demantk´e et al., 2011), custom descriptors for vertical cylindrical shapes (Monnier et al., 2012) and the height distribution in a vertical column around the point (Weinmann et…”