Tree Species Classification Based on 3D Bark Texture Analysis
Résumé
TerrestrialLaserScanning(TLS)techniqueistodaywidelyusedingroundplotstoacquire3Dpointcloudsfromwhichforestin-ventoryattributesarecalculated.Inthecaseofmixedplantingswherethe3Dpointcloudscontaindatafromseveraldifferenttreespecies,itisimportanttobeabletoautomaticallyrecognizethetreespeciesinordertoanalyzethedataofeachofthespeciesseparately.AlthoughautomatictreespeciesrecognitionfromTLSdataisanimportantproblem,ithasreceivedverylittleattentionfromthescientificcommunity.InthispaperweproposeamethodforclassifyingfivedifferenttreespeciesusingTLSdata.Ourmethodisbasedontheanalysisofthe3Dgeometrictextureofthebarkinordertocomputeroughnessmeasuresandshapecharacteris-ticsthatarefedasinputtoaRandomForestclassifiertoclassifythetreespecies.Themethodhasbeenevaluatedonatestsetcomposedof265samples(53samplesofeachofthe5species)andtheresultsobtainedareveryencouraging.
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...