ALOS影像数据土地覆盖分类及景观特征研究.pdf

216 2012，48(24) ComputerEngineeringandApplications计算机工程与应用 ALOS影像数据土地覆盖分类及景观特征研究 张 斌 ，王继尧 ，吕一河 ，曹 艺 ZHANG Bin， NG Jiyao LVYihe，CAOYi 1．中国矿业大学 资源与地球科学学院，江苏 徐州 221116 2-中国科学院 生态环境研究中心，北京 100085 1．SchoolofResourcesandEarthSciences，ChinaUniversityofM iningandTechnology,Xuzhou，Jiangsu221116，China 2．ResearchCenterforEco—EnvironmentalSciences，ChineseAcademyofSciences，Beijing100085，China ZHANG Bin，WANG Jiyao，LVYihe，eta1．Studyonlandcoverclassificationandlandscapecharacteristicsof ALOSimages．ComputerEngineeringandApplications，2012，48(24)：216-221． Abstract：LandcoverclassificationisprocessedbyM ahalanobisDistancemethod．maximum likelihoodmethod． and supportvectormachinemethodwithconfusionmatrixassessingtheclassificationaccuracy．Theresultsshow thatthemaximum likelihoodmemodandsupportvectormachinemethodareeffective．M aximum likelihoodmeth． odiStakenasanexampleforlandcoverclassificationperformanceanalysisbyintroducingtheNormalizedDifrer_ enceVegetationIndex(NDVI)andGLCM texturefeatures．ResultsrevealthattheclassificationaccuracyiSim． provedtovaryingdegreeswhenNDVI，CONTRAST，MEAN areincorporatedinclassificationprocess．Further combinationofthethreefeautreswiththeoriginalbandrepresentsthehighestclassificationaccuracy．Thegeneral landscapepattem characteristicsarequantifiedbasedon classificationresultsand landscapemetrics．Theresults show thatcroplandandhumansettlementsaredominatingthelandscapewithanarearatioof82％，whichimplies highhumanpressureandecologicalriskinthelandscape．Itis，therefore，necessary tostrengthenlanduseplanning andmanagementinthedesert—oasistransitionalareainthemiddlereachesoftheHeiheRiver~Specifically,concrete measuresareurgentlyneededto slow downandregulatecroplandandresidentialareaexpansionwith improved landuseefficiency．Atthesametime．ecologicalconservation andrehabilitationneedtobeenhancedtoincrease landscaperesiliencetodisturbances． Keywords：ALOS：landcover；classification；landscapepattern 摘 要：通过马氏距离法、最大似然法、支持向量机三种途径对土地覆盖进行分类，以混淆矩阵对分类结