@Article{SousaTeiSilAndBra:2010:AvClBa,
author = "Sousa, Beatriz Fernandes Simplicio and Teixeira, Adunias dos
Santos and Silva, Francisco de Assis Tavares Ferreira da and
Andrade, Eunice Maia de and Braga, Arthur Pl{\'{\i}}nio de
Souza",
affiliation = "{} and {} and {Instituto Nacional de Pesquisas Espaciais (INPE)}",
title = "Avalia{\c{c}}{\~a}o de classificadores baseados em aprendizado
de m{\'a}quina para a classifica{\c{c}}{\~a}o do uso e
cobertura da terra no bioma caatinga",
journal = "Revista Brasileira de Cartografia",
year = "2010",
volume = "62",
number = "2 , setembro 2010.",
pages = "edi{\c{c}}{\~a}o Especial",
month = "set.",
keywords = "Intelig{\^e}ncia Artificial, Semi-{\'a}rido,
Classifica{\c{c}}{\~a}o de Imagens de Sat{\'e}lite, Artificial
Intelligence, Semi Arid, Satellite Image Classification.",
abstract = "O manejo adequado dos recursos naturais em ambientes fr{\'a}geis,
como o da Caatinga, requer o conhecimento de suas propriedades e
distribui{\c{c}}{\~a}o espacial. Nesse contexto, o trabalho tem
por objetivo avaliar o desempenho de dois algoritmos baseados em
aprendizado de m{\'a}quina (Multi Layer Perceptron (MLP) e o
Support Vector Machine (SVM)) e do m{\'e}todo da M{\'a}xima
Verossimilhan{\c{c}}a na classifica{\c{c}}{\~a}o do uso e
cobertura da terra no bioma Caatinga. Para o experimento, foi
utilizada uma imagem do sat{\'e}lite LANDSAT-5/TM contendo a
{\'a}rea de estudo localizada no munic{\'{\i}}pio de Iguatu-CE
e definidas as classes de cobertura da terra, a saber:
antropiza{\c{c}}{\~a}o por agricultura (APA), outros tipos de
antropiza{\c{c}}{\~a}o (OTA), {\'a}gua, caatinga herb{\'a}cea
arbustiva (CHA) e caatinga arb{\'o}rea densa (CAD). O desempenho
dos m{\'e}todos foi analisado atrav{\'e}s dos coeficientes de
Exatid{\~a}o Global (EG), Exatid{\~a}o Espec{\'{\i}}fica (EE)
e Kappa (K) calculados a partir dos dados da matriz de
confus{\~a}o correspondente {\`a} verdade terrestre. Os valores
do coeficiente de EG foram de: 86,03%, 82,14% e 81,2% e K de:
0,77, 0,76 e 0,75 nos m{\'e}todos SVM, MLP e M{\'a}xima
Verossimilhan{\c{c}}a, respectivamente. Os valores de EE foram
superiores a 70% para todos os classificadores testados. Os
resultados obtidos demonstram que os m{\'e}todos SVM e MLP
est{\~a}o aptos {\`a} classifica{\c{c}}{\~a}o dos padr{\~o}es
propostos, j{\'a} que apresentaram resultados semelhantes ao
m{\'e}todo tradicional da M{\'a}xima Verossimilhan{\c{c}}a.
Por{\'e}m, estes classificadores podem consumir mais tempo na
etapa de defini{\c{c}}{\~a}o dos par{\^a}metros da rede e de
processamento.ABSTRACT Proper management of natural resources in
fragile environments, such as the Caatinga, requires knowledge of
their properties and spatial distribution. In this context, the
study aims at evaluating the performance of two algorithms based
on machine learning (Multi Layer Perceptron (MLP) and Support
Vector Machine (SVM)) and the Maximum Proper management of natural
resources in fragile environments, such as the Caatinga, requires
knowledge of their properties and spatial distribution. In this
context, the study aims at evaluating the performance of two
algorithms based on machine learning (Multi Layer Perceptron (MLP)
and Support Vector Machine (SVM)) and the Maximum Likelihood
method to classify land use and land cover in the Caatinga biome.
For the experiment, it was used a satellite image of LANDSAT-5/TM
containing the study area located in the municipality of
Iguatu-CE, and classes of land cover, namely: anthropized by
agriculture, other types of anthropized, water, herbaceous shrub
savanna (CHA ) and dense arboreal savanna (CAD) were defined. The
performance of the methods was analyzed by the coefficient of
Global Accuracy (EG), Accuracy Specific (EE) and Kappa (K)
coefficient calculated with data taken from the confusion matrix
corresponding to ground truth. The coefficient of EG were: 86.03%,
82.14% and 81.2% and K: 0.77, 0.76 and 0.75 in the methods SVM,
MLP and maximum likelihood respectively. EE values were above 70%
for all classifiers tested. The results have shown that SVM and
MLP methods are suited to the classification of the proposed
standards, as it showed similar results to the traditional method
of maximum likelihood. However, these methods are more time
consuming in the stage of defining the parameters of the network
and may require more computation power during stage of
processing.",
issn = "0560-4613 and 1808-0936",
language = "pt",
targetfile = "62_ESPECIAL_02_6.pdf",
urlaccessdate = "23 maio 2024"
}