Comparative Analysis of Triangulation Libraries for Modeling Large Point Clouds from Land and Their Infrastructures

Luis Lopez-Fernandez,Damian Ortega-Terol,Pablo Rodriguez-Gonzalvez,Diego Gonzalez-Aguilera,David Hernandez-Lopez

doi:10.3390/infrastructures2010001

Luis Lopez-Fernandez, Damian Ortega-Terol + Show 3 more

Open Access

https://doi.org/10.3390/infrastructures2010001

Copy DOI

Abstract

Although the generation of large points clouds from geomatic techniques allows us to realize the topography and appearance of the terrain and its infrastructures (e.g., roads, bridges, buildings, etc.), all these 3D point clouds require an unavoidable step to be conveniently treated: the definition of the surface that connects these points in space through digital surface models (DSM). In addition, these point clouds sometimes have associated attributes and geometric constraints such as breaklines and/or exclusion areas, which require the implementation of efficient triangulation techniques that can cope with a high volume of information. This article aims to make a comparative analysis of different Delaunay triangulation libraries, open or with academic versions available for the scientific community, so that we can assess their suitability for the modeling of the territory and its infrastructures. The comparison was carried out from a two-fold perspective: (i) to analyze and compare the computational cost of the triangulation; (ii) to assess the geometric quality of the resulting meshes. The different techniques and libraries have been tested based on three different study cases and the corresponding large points clouds generated. The study has been useful to identify the limitations of the existing large point clouds triangulation libraries and to propose statistical variables that assess the geometric quality of the resulting DSM.

Highlights

The beginning of this century has been characterized by the advancement of geomatics techniques, from remote data collection to processing and visualization of geoinformation [1]
Technologies such as Light Detection and Ranging (LiDAR) [2] and the emergence of new sensors in photogrammetry [3] and remote sensing (e.g., Sentinel) [4] have led to a remarkable growth in the quality of geomatic products, in particular (i) the generation of 2D cartographic products in the form of maps and orthoimages [5]; (ii) the generation of 3D cartographic products in the form of digital surface models (DSM) [6], canopy height models (CHM) [7] and even digital city models (DCM) [8]; and (iii) monitoring and simulation based on the temporal analysis of these products [9]
All land-related disciplines have benefited, to a greater or lesser extent, from these developments, but especially civil and surveying engineering, since they allow for land characterization along with their associated infrastructures

Summary

Introduction

The beginning of this century has been characterized by the advancement of geomatics techniques, from remote data collection to processing and visualization of geoinformation [1]. Modern photogrammetry and dense matching techniques [12] provide better resolution than LiDAR (one depth or one elevation is generated per pixel), they require a high photogrammetric computational cost. In any case, both LiDAR and photogrammetric point clouds require the extraction of relevant and useful information beyond their own point cloud. The triangulation of point clouds in the TIN format becomes an unavoidable step that requires a large computational cost, especially with large point clouds [14]

Objectives

Methods

Findings

Conclusion