Classification of Australian Native Forest Species Using Hyperspectral Remote Sensing and Machine-Learning Classification Algorithms

Mapping forest species is highly relevant for many ecological and forestry applications. In Australia, the classification of native forest species using remote sensing data remains a particular challenge since there are many eucalyptus species that belong to the same genus and, thus, exhibit similar...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE journal of selected topics in applied earth observations and remote sensing Jg. 7; H. 6; S. 2481 - 2489
Hauptverfasser: Shang, Xiao, Chisholm, Laurie A.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Piscataway IEEE 01.06.2014
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Accuracy
Algorithms
Canopies
Classification
Communities
Comparative studies
Forestry
Indigenous species
Leaves
Predation
Radio frequency
Remote sensing
Support vector machines
Vegetation
Vegetation mapping
Australia
ISSN:1939-1404, 2151-1535
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Mapping forest species is highly relevant for many ecological and forestry applications. In Australia, the classification of native forest species using remote sensing data remains a particular challenge since there are many eucalyptus species that belong to the same genus and, thus, exhibit similar biophysical characteristics. This study assessed the potential of using hyperspectral remote sensing data and state-of-the-art machine-learning classification algorithms to classify Australian forest species at the leaf, canopy and community levels in Beecroft Peninsula, NSW, Australia. Spectral reflectance was acquired from an ASD spectrometer and airborne Hymap imagery for seven native forest species over an Australian eucalyptus forest. Three machine-learning classification algorithms: Support Vector Machine (SVM), AdaBoost and Random Forest (RF) were applied to classify the species. A comparative study was carried out between machine-learning classification algorithms and Linear Discriminant Analysis (LDA). The classification results show that all machine-leaning classification algorithms significantly improve the results produced by LDA. At the leaf level, RF achieved the best classification accuracy (94.7%), and SVM outperformed the other algorithms at both the canopy (84.5%) and community levels (75.5%). This study demonstrates that hyperspectral remote sensing and machine-learning classification has substantial potential for the classification of Australian native forest species.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1939-1404
2151-1535
DOI:10.1109/JSTARS.2013.2282166