基于Hyperion高光谱数据和随机森林方法的岩性分类与分析-岩石学报.PDF

基于Hyperion高光谱数据和随机森林方法的岩性分类与分析-岩石学报.PDF

10000569/2018/034(07)218188 ActaPetrologicaSinica 岩石学报 基于Hyperion高光谱数据和随机森林方法的岩性 分类与分析 1 1 1 2 柯元楚 史忠奎 李培军  张西雅 1 1 1 2 KEYuanChu,SHIZhongKui,LIPeiJun andZHANGXiYa 1北京大学地球与空间科学学院遥感与地理信息系统研究所,北京 100871 2中国气象局北京城市气象研究所,北京 100089 1InstituteofRemoteSensingandGIS,SchoolofEarthandSpaceSciences,PekingUniversity,Beijing100871,China 2InstituteofUrbanMeteorology,ChinaMeteorologicalAdministration,Beijing100089,China 20180120收稿,20180422改回 KeYC,ShiZK,LiPJandZhangXY2018LithologicalclassificationandanalysisusingHyperionhyperspectraldataand RandomForestmethodActaPetrologicaSinica,34(7):2181-2188 Abstract  Explorationofnewmethodsforlithologicalmappingusinghyperspectraldataisoneofcriticalneedsinthefieldof geologicalremotesensingapplicationsInthispaper,randomforestmethodwasusedtoclassifyEO1Hyperionhyperspectraldatafor lithologicalmappingoveraportionofKalpin,NorthwestTarim,Xinjiang,andtherelevantproblemswereanalyzedThelithological classificationwascarriedoutbyusingspectralfeaturesaloneandusingbothspectralfeaturesandfirstderivativesTheimportanceof differentfeaturestolithologicalclassificationwascalculatedandanalyzedTheRandomForestmethodwascomparedwithspectralangle mapper(SAM),acommonlyusedclassificationmethodTheresultsshowthattheRandomForestmethodachievedasignificant improvementinaccuracyuponlithologicalclassificationcomparedwithSAMThus,theadoptedmethodisafeasiblelithological classificationmethodAccordingtotheorderingoffeatureimportance,blueandgreen,shortwaveinfraredbandsandtheir correspondingfirstderivativescontributemoretolithologicalclassificationofsedimentaryrocksinthestudyarea Keywords  Hyperspectraldata;RandomForest;Spectralanglemapper;Lithologicalclassification 摘 要  探索利用高光谱数据的岩性填图新方法是遥感地质应用领域的重要需求之一。本文运用随机森林方法和 EO1 Hyperion高光谱数据,对新疆塔里木西北部柯坪地区的局部区域进行岩性分类,并对相关问题进行分析。分别利用光谱特征 以及加入光谱一阶导数特征进行岩性分类,并对不同特征对岩性分类的重要性进行分析,同时与现有的基于光谱角制图方法 (SAM)进行比较。结果表明,与SAM方法相

文档评论(0)

1亿VIP精品文档

相关文档