@InProceedings{SapucciNegr:2017:PrClSe,
author = "Sapucci, Gabriela Ribeiro and Negri, Rog{\'e}rio Galante",
title = "Proposta de Classificadores Semissupervisionados baseados em
Rotula{\c{c}}{\~a}o de Agrupamentos via Dist{\^a}ncias
Estoc{\'a}sticas",
booktitle = "Anais...",
year = "2017",
editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz
Eduardo Oliveira e Cruz de",
pages = "7694--7700",
organization = "Simp{\'o}sio Brasileiro de Sensoriamento Remoto, 18. (SBSR)",
publisher = "Instituto Nacional de Pesquisas Espaciais (INPE)",
address = "S{\~a}o Jos{\'e} dos Campos",
abstract = "Remote sensing image classification is one of the most important
applications of Pattern Recognition in environmental studies.
Image classification methods generally have supervised learning or
unsupervised. As supervised learning methods perform sorting by
means of a function or decision rule modeled through information
provided in advance, the quality of the results is directly
related to the quality of the set of training standards, which
doesn''t always guarantee quality results. Unsupervised learning,
in turn, build your knowledge in function of analogies observed
about the data, which can be a complex task. Alternatively, the
semi-supervised learning aims to deal with the weaknesses of both
paradigms, by combining concepts of learning with and without
supervision. In this context, this research project proposes the
formalization and implementation of two methods of semi-supervised
classification, which combines classic tools in the area of
pattern recognition: the Hierarchical Divisive Algorithms,
\$K\$-Means and stochastic distances. From a set of groups,
defined by the combination of Hierarchical Divisive Algorithm and
\$K\$-Means and another defined only by \$K\$-Means, through
unsupervised learning, stochastic distances are used for labeling
of each of these groups. Through case studies on the use and
classification of ground cover around the Tapaj{\'o}s National
Forest, the quality of the results obtained according to the Kappa
coefficient was analyzed and the proposed methods were compared
with other classification methods already known in the
literature.",
conference-location = "Santos",
conference-year = "28-31 maio 2017",
isbn = "978-85-17-00088-1",
label = "60162",
language = "pt",
organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
ibi = "8JMKD3MGP6W34M/3PSMG97",
url = "http://urlib.net/ibi/8JMKD3MGP6W34M/3PSMG97",
targetfile = "60162.pdf",
type = "Processamento de imagens",
urlaccessdate = "27 abr. 2024"
}