Classifying spatially heterogeneous wetland communities using machine learning algorithms and spectral and textural features

Título traducido de la contribución: Clasificación de comunidades de humedales espacialmente heterogéneas utilizando algoritmos de aprendizaje automático y características espectrales y texturales.

Zoltan Szantoi, Francisco J. Escobedo, Amr Abd-Elrahman, Leonard Pearlstine, Bon Dewitt, Scot Smith

Resultado de la investigación: Contribución a RevistaArtículo

15 Citas (Scopus)

Resumen

La cartografía de los humedales (pantanos vs. pantanos vs. tierras altas) es una aplicación común de la teledetección; sin embargo, es más difícil discriminar entre comunidades de agua dulce similares, como graminoides/bordes, a partir de imágenes obtenidas por teledetección. La mayor parte de esta actividad se ha realizado utilizando imágenes de resolución media a baja. Existen pocos estudios que utilicen imágenes de alta resolución espacial y algoritmos de clasificación de imágenes de aprendizaje automático para cartografiar comunidades de plantas de humedales heterogéneas. Este estudio aborda este vacío analizando si los clasificadores de aprendizaje automático tales como los árboles de decisión (DT) y las redes neurales artificiales (ANN) pueden clasificar con precisión las comunidades de graminoides/sedgecomunidades usando imágenes aéreas de alta resolución y datos de textura de imagen en el Parque Nacional Everglades, Florida, Además de las bandas espectrales, se analizaron el índice de diferenciación de vegetación normalizado y las características de textura de primer y segundo orden derivadas de la banda infrarroja cercana. Se evaluaron las precisiones de los clasificadores utilizando tablas de confusión y los coeficientes kappa calculados de los mapas resultantes. Los resultados indicaron que un algoritmo ANN (perceptrón multicapa basado en la retropropagación) producía una precisión estadísticamente significativa (82,04%) mayor que el algoritmo DT (QUEST) (80,48%) o el clasificador de máxima verosimilitud (80,56%) (α<0,05). Los resultados muestran que el uso de múltiples tamaños de ventana proporcionó los mejores resultados. Las características de textura de primer orden también proporcionaron ventajas computacionales y resultados que no fueron significativamente diferentes de las características de textura de segundo orden.
Idioma originalEnglish (US)
Páginas (desde-hasta)262
Número de páginas1
PublicaciónEnvironmental Monitoring and Assessment
DOI
EstadoPublished - may 1 2015
Publicado de forma externa

Huella dactilar

Wetlands
Learning algorithms
Learning systems
Classifiers
imagery
wetland
Neural networks
Image texture
artificial neural network
Backpropagation algorithms
Image classification
Maximum likelihood
Remote sensing
sedge
image classification
Antennas
Infrared radiation
swamp
NDVI
void

Citar esto

@article{497f29fe660345b89198005363db374a,
title = "Classifying spatially heterogeneous wetland communities using machine learning algorithms and spectral and textural features",
abstract = "Mapping of wetlands (marsh vs. swamp vs. upland) is a common remote sensing application.Yet, discriminating between similar freshwater communities such as graminoid/sedge fromremotely sensed imagery is more difficult. Most of this activity has been performed using medium to low resolution imagery. There are only a few studies using highspatial resolutionimagery and machine learning image classification algorithms for mapping heterogeneouswetland plantcommunities. This study addresses this void by analyzing whether machine learning classifierssuch as decisiontrees (DT) and artificial neural networks (ANN) can accurately classify graminoid/sedgecommunities usinghigh resolution aerial imagery and image texture data in the Everglades National Park, Florida.In addition tospectral bands, the normalized difference vegetation index, and first- and second-order texturefeatures derivedfrom the near-infrared band were analyzed. Classifier accuracies were assessed using confusiontablesand the calculated kappa coefficients of the resulting maps. The results indicated that an ANN(multilayerperceptron based on backpropagation) algorithm produced a statistically significantly higheraccuracy(82.04{\%}) than the DT (QUEST) algorithm (80.48{\%}) or the maximum likelihood (80.56{\%})classifier (α",
author = "Zoltan Szantoi and Escobedo, {Francisco J.} and Amr Abd-Elrahman and Leonard Pearlstine and Bon Dewitt and Scot Smith",
year = "2015",
month = "5",
day = "1",
doi = "10.1007/s10661-015-4426-5",
language = "English (US)",
pages = "262",
journal = "Environmental Monitoring and Assessment",
issn = "0167-6369",
publisher = "Springer Netherlands",

}

Classifying spatially heterogeneous wetland communities using machine learning algorithms and spectral and textural features. / Szantoi, Zoltan; Escobedo, Francisco J.; Abd-Elrahman, Amr; Pearlstine, Leonard; Dewitt, Bon; Smith, Scot.

En: Environmental Monitoring and Assessment, 01.05.2015, p. 262.

Resultado de la investigación: Contribución a RevistaArtículo

TY - JOUR

T1 - Classifying spatially heterogeneous wetland communities using machine learning algorithms and spectral and textural features

AU - Szantoi, Zoltan

AU - Escobedo, Francisco J.

AU - Abd-Elrahman, Amr

AU - Pearlstine, Leonard

AU - Dewitt, Bon

AU - Smith, Scot

PY - 2015/5/1

Y1 - 2015/5/1

N2 - Mapping of wetlands (marsh vs. swamp vs. upland) is a common remote sensing application.Yet, discriminating between similar freshwater communities such as graminoid/sedge fromremotely sensed imagery is more difficult. Most of this activity has been performed using medium to low resolution imagery. There are only a few studies using highspatial resolutionimagery and machine learning image classification algorithms for mapping heterogeneouswetland plantcommunities. This study addresses this void by analyzing whether machine learning classifierssuch as decisiontrees (DT) and artificial neural networks (ANN) can accurately classify graminoid/sedgecommunities usinghigh resolution aerial imagery and image texture data in the Everglades National Park, Florida.In addition tospectral bands, the normalized difference vegetation index, and first- and second-order texturefeatures derivedfrom the near-infrared band were analyzed. Classifier accuracies were assessed using confusiontablesand the calculated kappa coefficients of the resulting maps. The results indicated that an ANN(multilayerperceptron based on backpropagation) algorithm produced a statistically significantly higheraccuracy(82.04%) than the DT (QUEST) algorithm (80.48%) or the maximum likelihood (80.56%)classifier (α

AB - Mapping of wetlands (marsh vs. swamp vs. upland) is a common remote sensing application.Yet, discriminating between similar freshwater communities such as graminoid/sedge fromremotely sensed imagery is more difficult. Most of this activity has been performed using medium to low resolution imagery. There are only a few studies using highspatial resolutionimagery and machine learning image classification algorithms for mapping heterogeneouswetland plantcommunities. This study addresses this void by analyzing whether machine learning classifierssuch as decisiontrees (DT) and artificial neural networks (ANN) can accurately classify graminoid/sedgecommunities usinghigh resolution aerial imagery and image texture data in the Everglades National Park, Florida.In addition tospectral bands, the normalized difference vegetation index, and first- and second-order texturefeatures derivedfrom the near-infrared band were analyzed. Classifier accuracies were assessed using confusiontablesand the calculated kappa coefficients of the resulting maps. The results indicated that an ANN(multilayerperceptron based on backpropagation) algorithm produced a statistically significantly higheraccuracy(82.04%) than the DT (QUEST) algorithm (80.48%) or the maximum likelihood (80.56%)classifier (α

U2 - 10.1007/s10661-015-4426-5

DO - 10.1007/s10661-015-4426-5

M3 - Article

C2 - 25893753

SP - 262

JO - Environmental Monitoring and Assessment

JF - Environmental Monitoring and Assessment

SN - 0167-6369

ER -