Tree species classification in heterogeneous forest with hyperspectral data from unmanned aerial vehicle

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27350%2F25%3A10258854" target="_blank" >RIV/61989100:27350/25:10258854 - isvavai.cz</a>

Výsledek na webu

<a href="https://journals.pan.pl/dlibra/publication/154152/edition/137225/content" target="_blank" >https://journals.pan.pl/dlibra/publication/154152/edition/137225/content</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.24425/agg.2025.154152" target="_blank" >10.24425/agg.2025.154152</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Tree species classification in heterogeneous forest with hyperspectral data from unmanned aerial vehicle

Popis výsledku v původním jazyce

Hyperspectral data obtained from unmanned aerial vehicles (UAVs) provide high spectral and spatial resolution, bringing potential for an accurate identification of individual tree species. The aim of this paper was to thoroughly investigate the possibilities of forest tree classification using hyperspectral data taken by the Resonon Pika L camera. Both hyperspectral and ground reference data were collected for a heterogenous forest in the Czech Republic. Standard processing methods (radiometric, atmospheric, and geometric corrections) were applied, followed by testing the methods for reduction (spectral resampling, MNF, PCA) and PPI. The classification phase consists of both unsupervised and supervised approaches, including Maximum Likelihood, Mahalanobis Distance, Spectral Angle Mapper, Minimum Distance, Random Forest, Extra Trees, Support Vector Machines, and Naive Bayes. Within the classical classifiers, the best results were achieved using the Maximum Likelihood classifier. In terms of machine learning algorithms, the best performing classifiers were Random Forest and Extra Trees. The use of Pika L camera in forestry classification is so far minimal, therefore the results can be helpful in potential utilization of this type of camera. The findings of this research not only contribute to a better understanding of UAV-based hyperspectral remote sensing for tree classifications but also provide practical insights andrecommendations for improvement

Název v anglickém jazyce

Tree species classification in heterogeneous forest with hyperspectral data from unmanned aerial vehicle

Popis výsledku anglicky

Hyperspectral data obtained from unmanned aerial vehicles (UAVs) provide high spectral and spatial resolution, bringing potential for an accurate identification of individual tree species. The aim of this paper was to thoroughly investigate the possibilities of forest tree classification using hyperspectral data taken by the Resonon Pika L camera. Both hyperspectral and ground reference data were collected for a heterogenous forest in the Czech Republic. Standard processing methods (radiometric, atmospheric, and geometric corrections) were applied, followed by testing the methods for reduction (spectral resampling, MNF, PCA) and PPI. The classification phase consists of both unsupervised and supervised approaches, including Maximum Likelihood, Mahalanobis Distance, Spectral Angle Mapper, Minimum Distance, Random Forest, Extra Trees, Support Vector Machines, and Naive Bayes. Within the classical classifiers, the best results were achieved using the Maximum Likelihood classifier. In terms of machine learning algorithms, the best performing classifiers were Random Forest and Extra Trees. The use of Pika L camera in forestry classification is so far minimal, therefore the results can be helpful in potential utilization of this type of camera. The findings of this research not only contribute to a better understanding of UAV-based hyperspectral remote sensing for tree classifications but also provide practical insights andrecommendations for improvement

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

20705 - Remote sensing

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Advances in Geodesy and Geoinformation

ISSN

2720-7242

e-ISSN

—

Svazek periodika

74

Číslo periodika v rámci svazku

2

Stát vydavatele periodika

PL - Polská republika

Počet stran výsledku

13

Strana od-do

nestránkováno

Kód UT WoS článku

001628126500001

EID výsledku v databázi Scopus

—