Image Classification using Neural Networks and Random Forest for a Citizen Science Project

Abstract

See, stats, and : https : / / www . researchgate . net / publication / 314033623 Image Random Conference CITATIONS 0 READS 36 3 , including : Marilyn National , Brazil 4 SEE Adenilson Departamento 11 SEE All . The . Abstract Recently , a citizen science project called ForestWatchers [ 1 ] was launched in order to involve the citizens without experience in the monitoring of deforestation . Through a Web interface , volunteers from around the world are invited to review MODIS images of forest regions . Considering the large area worldwide covered by tropical forest , it is essential to use a fast classifier that meetsadoubleobjective:thepixelmappingintotwoclasses('Forest'and'Non-forest')andtheselectionofpixelstobesenttovolunteersforinspection.ThisprojectinvestigatestheusethedeforestationpatternclassificationintheBrazilianAmazonusingMODISimages.Deforestationmapsweregeneratedfromvari-oussizes,fromdifferentareasofthestateofRondonia.TheresultswerevalidatedwiththeresultsofPRODESproject,whichannuallyevaluatesdeforestationintheBrazilianAmazon.