Article Dans Une Revue Remote Sensing Année : 2024

A New Open-Source Software to Help Design Models for Automatic 3D Point Cloud Classification in Coastal Studies

Résumé

This study introduces a new software, cLASpy_T, that helps design models for the automatic 3D point cloud classification of coastal environments. This software is based on machine learning algorithms from the scikit-learn library and can classify point clouds derived from LiDAR or photogrammetry. Input data can be imported via CSV or LAS files, providing a 3D point cloud, enhanced with geometric features or spectral information, such as colors from orthophotos or hyperspectral data. cLASpy_T lets the user run three supervised machine learning algorithms from the scikit-learn API to build automatic classification models: RandomForestClassifier, GradientBoostingClassifier and MLPClassifier. This work presents the general method for classification model design using cLASpy_T and the software's complete workflow with an example of photogrammetry point cloud classification. Four photogrammetric models of a coastal dike were acquired on four different dates, in 2021. The aim is to classify each point according to whether it belongs to the 'sand' class of the beach, the 'rock' class of the riprap, or the 'block' class of the concrete blocks. This case study highlights the importance of adjusting algorithm parameters, selecting features, and the large number of tests necessary to design a classification model that can be generalized and used in production.

Fichier principal
Vignette du fichier
remotesensing-16-02891.pdf (49.14 Mo) Télécharger le fichier
Origine Fichiers éditeurs autorisés sur une archive ouverte
Licence

Dates et versions

insu-04725532 , version 1 (08-10-2024)

Licence

Identifiants

Citer

X. Pellerin Le Bas, Laurent Froideval, Adan Mouko, Christophe Conessa, Laurent Benoit, et al.. A New Open-Source Software to Help Design Models for Automatic 3D Point Cloud Classification in Coastal Studies. Remote Sensing, 2024, 16, ⟨10.3390/rs16162891⟩. ⟨insu-04725532⟩
107 Consultations
65 Téléchargements

Altmetric

Partager

  • More