<?xml version="1.0" encoding="ISO-8859-1"?>
<metadatalist>
	<metadata ReferenceType="Conference Proceedings">
		<site>marte2.sid.inpe.br 802</site>
		<holdercode>{isadg {BR SPINPE} ibi 8JMKD3MGPCW/3DT298S}</holdercode>
		<identifier>3ERPFQRTRW34M/3E7GJ5D</identifier>
		<repository>dpi.inpe.br/marte2/2013/05.29.00.24.56</repository>
		<lastupdate>2013:05.29.00.24.56 dpi.inpe.br/marte2/2013/05.17.15.03.06 administrator</lastupdate>
		<metadatarepository>dpi.inpe.br/marte2/2013/05.29.00.24.57</metadatarepository>
		<metadatalastupdate>2018:06.06.03.33.15 dpi.inpe.br/marte2/2013/05.17.15.03.06 administrator {D 2013}</metadatalastupdate>
		<isbn>978-85-17-00066-9 (Internet)</isbn>
		<isbn>978-85-17-00065-2 (DVD)</isbn>
		<label>1102</label>
		<citationkey>AbdallaVolo:2013:EsCoDi</citationkey>
		<title>Estudo da configuração de diferentes arquiteturas de redes neurais artificiais MLP para classificação de imagens ópticas</title>
		<format>DVD, Internet.</format>
		<year>2013</year>
		<secondarytype>PRE CN</secondarytype>
		<numberoffiles>1</numberoffiles>
		<size>183 KiB</size>
		<author>Abdalla, Livia dos Santos,</author>
		<author>Volotão, Carlos Frederico de Sá,</author>
		<electronicmailaddress>abdalla.livia@gmail.com</electronicmailaddress>
		<editor>Epiphanio, José Carlos Neves,</editor>
		<editor>Galvão, Lênio Soares,</editor>
		<conferencename>Simpósio Brasileiro de Sensoriamento Remoto, 16 (SBSR)</conferencename>
		<conferencelocation>Foz do Iguaçu</conferencelocation>
		<date>13-18 abr. 2013</date>
		<publisher>Instituto Nacional de Pesquisas Espaciais (INPE)</publisher>
		<publisheraddress>São José dos Campos</publisheraddress>
		<pages>8200-8207</pages>
		<booktitle>Anais</booktitle>
		<organization>Instituto Nacional de Pesquisas Espaciais (INPE)</organization>
		<transferableflag>1</transferableflag>
		<abstract>Artificial neural networks (ANN) can be used to produce several products, including remote sensing classification images.  ANN may not beat the performance of traditional classification methods, but it is unique in the sense that: 1) it is not dependent on the prior knowledge of statistical model of data; and 2) it makes it possible to add unusual information by the configuration of parameters, input, hidden and output layers. Motivated by the ability to add different levels of information, including spatial and non-spatial data (e.g., Digital Terrain Models, time, date, a given classification or a segmented image), and comparing to classical methods of classification, this work test the use of ANN for image classification.  Despite this capability, this work aims to compare the plain classification ability, by means of kappa values and training sets as a reference example when there is no ground truth available.  Providing a fare alternative for image classification, the advantages of the potential enhancements are to be studied in future papers. This study explores simple architectures of MLP to identify common themes of land cover and uses, and is based on HRG/SPOT5 images. The results using  kappa was 91% indicating that the RNA has achieved a good index of training.</abstract>
		<area>SRE</area>
		<type>Processamento de Imagens</type>
		<language>pt</language>
		<targetfile>p1102.pdf</targetfile>
		<usergroup>administrator</usergroup>
		<usergroup>banon</usergroup>
		<visibility>shown</visibility>
		<mirrorrepository>urlib.net/www/2011/03.29.20.55</mirrorrepository>
		<nexthigherunit>3ERPFQRTRW34M/3E7G88S</nexthigherunit>
		<citingitemlist>dpi.inpe.br/marte2/2013/05.28.22.25 3</citingitemlist>
		<hostcollection>dpi.inpe.br/marte2/2013/05.17.15.03.06</hostcollection>
		<username>wanderf@dsr.inpe.br</username>
		<lasthostcollection>dpi.inpe.br/marte2/2013/05.17.15.03.06</lasthostcollection>
		<url>http://marte2.sid.inpe.br/rep-/dpi.inpe.br/marte2/2013/05.29.00.24.56</url>
	</metadata>
</metadatalist>