多源遥感汇报ppt课件.pptx

影像与激光雷达数据的城市地物信息提取21.研究区及数据2.研究方法2.1数据预处理获取地物类型最佳分割尺度：如图5所示的分割质量曲线图，其中建设用地包括建筑物、道路、空地；林地包括树木及草地。从图中可看出随着分割尺度的增大，p值减少，r值增加，过分割现象减少，欠分割现象增多，而m2为这两种度量的加权和，不存在偏向性。因此最佳分割尺度一般取m2最大值处，若m2近似则选取p值较大处，得出各类地物类型的最佳分割尺度。树木与草地为33，道路与空地为45，阴影为18。2.3对象分类
特征选择后采用两步法提取城市地物，首先采用多分类器组合(K近邻算法、神经元网络及SVM_RBF)对城市地物进行分类，得到如图6所示的分类结果。由于城市地区存在高层建筑及树木，分类受阴影的影响较大，因此定本文在得出初步分类结果后采用易康5.0将阴影区域进行分割，并基于规则将阴影区域依次分为：阴影下植被、阴影下建筑物、阴影下道路、阴影下空地、纯阴影5类。
图7为A区最终分类结果，利用混淆矩阵法进行分类精度评价，总体精度为93.1%，如表2所示

2.4方法检验3结束语MonitoringcoalfiresinDatongcoalfieldusingmulti-sourceremotesensingdataAbstract:TheDatongcoalfieldislocatedinnorthernShanxiProvince,Chinaapproximatelyfrom39°52′to40°10′northlatitudeand112°49′32″to113°9′30″westlongitude。Ithasacomplexterrainwithanaveragealtitudeofmorethan1200m.ThestudiedareaislocatedintheMajiliangminingareaCoalfiresmainlyoccurintheNo.2Jurassiccoalseamsintheminingarea,30mto130mbelowthesurface,withanaveragethicknessofabout1m。3Multi-sourceremotesensingmonitoring3.1CoalfiremonitoringusingLandsatthermalinfraredband(a)R(Band5),G(Band4)andB(Band3)compositeimageswith15mspatialresolution(b)−(f)temperaturedistributionretrievalbyLandsatthermalinfraredbandwith60mspatialresolution3.2CoalfiremonitoringusingUAVBycontrast,comparedwithsatelliteremotesensing,UAVhasmanyadva-ntages.Itismoreflexibleintermsofworkingtimeandregion,hashigherspatialresolution(morethan0.2m),andiscosteffective.Itprovidesmuchbetterimagedataforgroundfissures.Fig4UAVimagewith0.2mspatialresolutionabletocaturegroundfissureincoalfireregionsAccordingtothetextureinformation,linearfeature,andbrightnessofthegroundfissures,aknowledgemodelwasestablishedtofacilitatetheautomaticextractionofgroundfissures.Thestepsareasfollows.1)Occurrence-basedvariance,cooccurrencebasedvariance,datarange,andcontrastfiltersareusedfortheUAVimagetoobtainthetextureinformationinthefissuredareabyamovingwindow.基于方差的发生与共生？
2)Principlecomponentanalysis(PCA)andFisherlineardiscrimination(线性判别分析)analysisarethenperformedtoextractthelinearfeaturesofthearea.
3)Agraylevelstatisti