Reducing bias in a triangulation-based variant of spatial clustering

@inproceedings{MICHALAK_RM2021,
 abstract = { Spatial clustering is a generic term for investigating geometric trends within geological objects. The first step requires finding homogenous subsets from among collected orientations with a known locality. Each observation receives a label identifying a specific subset. Although the spatial information is not involved in the labeling procedure, it is attached to the labeled observations at the end of the process. This study explores a triangulation-based variant of spatial clustering. The unit observations are Delaunay triangles sampled throughout a geologic surface, i.e., an interface separating two geological units. Because the labels are attached to geometric centers of Delaunay triangles, the resulting map is biased. Here, we present an outline of a method to solve the representativeness problem by creating a regular version of the proposed variant. This version creates a regular grid of points that are linked to Delaunay triangles by specific query functions implemented in the CGAL library. We present the applicability of the method for synthetic as well as real data from Kraków-Silesian Homocline. We argue that this data-consuming approach may be a promising tool for extracting structural information from faulted surfaces. For example, it may be helpful for the recognition of faults and their topology. However, we also highlight some fundamental limitations related to regional geology and the principles of computational geometry.
The study was supported by National Science Centre, Poland, 2020/37/N/ST10/02504. },
 author = { Michalak, Michał AND Kostur, Marcin AND Teper, Lesław AND Wellmann, Florian },
 booktitle = { 2021 RING Meeting },
 publisher = { ASGA },
 title = { Reducing bias in a triangulation-based variant of spatial clustering },
 year = { 2021 }
}