@Article{SotheLiesAlmeSchi:2017:AbClEs,
author = "Sothe, Camile and Liesenberg, Veraldo and Almeida, Cl{\'a}udia
Maria de and Schimalski, Marcos Benedito",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and
{Universidade do Estado de Santa Catarina (UDESC)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)} and {Universidade do
Estado de Santa Catarina (UDESC)}",
title = "Abordagens para classifica{\c{c}}{\~a}o do est{\'a}gio
sucessional da vegeta{\c{c}}{\~a}o do Parque Nacional de
S{\~a}o Joaquim empregando imagens Landsat-8 e Rapideye",
journal = "Boletim de Ci{\^e}ncias Geod{\'e}sicas",
year = "2017",
volume = "23",
number = "3",
pages = "389--404",
month = "jul./set.",
keywords = ": Sucess{\~a}o florestal, M{\'a}quinas de Vetor de Suporte,
Florestas Rand{\^o}micas, Atributos, Secondary forests, Support
Vector Machine, Random Forest, Features.",
abstract = "A classifica{\c{c}}{\~a}o remota dos diferentes est{\'a}dios
sucessionais da vegeta{\c{c}}{\~a}o ainda constitui um desafio
devido {\`a} similaridade espectral destas classes. Este artigo
tem o objetivo de avaliar o desempenho de imagens Landsat-8 e
RapidEye para a classifica{\c{c}}{\~a}o do est{\'a}dio
sucessional da vegeta{\c{c}}{\~a}o em um fragmento de Floresta
Ombr{\'o}fila Mista, localizado no Parque Nacional de S{\~a}o
Joaquim- SC. Para isto, tr{\^e}s grupos de vari{\'a}veis gerados
a partir de cada imagem foram avaliados, sendo: (1) composto
somente pelas bandas espectrais puras; (2) composto pelas
m{\'e}tricas texturais GLCM geradas a partir das bandas
espectrais; e (3) composto pelas vari{\'a}veis dos dois grupos
anteriores, al{\'e}m de dois {\'{\i}}ndices de
vegeta{\c{c}}{\~a}o no caso da imagem Landsat-8, e tr{\^e}s
{\'{\i}}ndices para a RapidEye. Cada grupo foi testado com os
classificadores florestas rand{\^o}micas (Random Forest- RF),
m{\'a}quinas de vetor de suporte (Support Vector Machine - SVM) e
m{\'a}xima verossimilhan{\c{c}}a (Maxver). Todos os experimentos
alcan{\c{c}}aram resultados satisfat{\'o}rios, com
{\'{\i}}ndice Kappa variando de 0,66 a 0,88 e acur{\'a}cia de
usu{\'a}rio e produtor superiores a 50%. O melhor resultado
alcan{\c{c}}ado foi com a imagem Landsat-8, grupo 3, associado ao
algoritmo RF. A medida de import{\^a}ncia das vari{\'a}veis
obtida com o algoritmo RF mostrou que as m{\'e}tricas texturais
m{\'e}dia, contraste e dissimilaridade destacaram-se na
classifica{\c{c}}{\~a}o para ambas as imagens. ABSTRACT: The
remote classification of the different vegetation successional
stages still represents a challenging task in face of the similar
spectral response of such classes. This paper is committed to
evaluate the performance of Landsat-8 and RapidEye images in the
classification of successional stages within a patch of Mixed
Ombrophilous Forest located in S{\~a}o Joaquim National Park,
Santa Catarina State, south of Brazil. Three variables dataset
extracted from each image were analyzed, namely; (1) one solely
consisting of the spectral bands themselves; (2) a second one
comprising GLCM-based texture measures derived from the spectral
bands; and (3) a third one containing these two datasets and
additionally two vegetation indices obtained from the Landsat 8
image and three vegetation indices from the RapidEye image. Each
dataset was subject to three classifiers: random forest (RF),
support vector machine (SVM), and maximum likelihood estimation
(MLE or maxver). All conducted experiments achieved satisfactory
results, with Kappa coefficients ranging from 0.66 to 0.88, and
both userīs and producerīs accuracies lying over 50%. The best
result was attained with the Landsat 8 image using the third
dataset and the RF classifier. The analysis of the variables
relevance with this classifier showed that the texture measures
mean, contrast and dissimilarity were decisive for the successful
classification of both images.",
doi = "10.1590/S1982-21702017000300026",
url = "http://dx.doi.org/10.1590/S1982-21702017000300026",
issn = "1413-4853",
language = "pt",
targetfile = "sothe_abordagens.pdf",
urlaccessdate = "27 abr. 2024"
}