期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2010
卷号：XXXVIII - Part 7A
页码：245-250
出版社：Copernicus Publications
摘要：The most extensive use of Remote Sensing data is in land cover/land use(LCLU)studies by means of automated imageclassification. Thegeneral objectiveof this researchisto developan automatic pixel-based classification methodologywith theaimto produceaRegional land use map congruent with the CORINELand Cover legend. Starting pointaredetailed ground data,alreadygathered fostering interoperability among several Regional bodies' DBsandhigh resolution multi-spectralIKONOSimagery.In the light of land mapping, there are two main features related to IKONOS imagery:lack of spectral information(4 spectral bands)andhighspectral variability(highspatialresolution). Thisresultsin problems in terms ofclassinformation extractionespeciallyusing pixel-based image classification methods in whichspatial information existing between a pixel and its neighboursis not used.To overcome thesedeficits,theuseof vegetation indexes (NDVI feature and TDVI masks) andtexture(GLCM and edge-densityfeatures) is investigated with respect to its impact onland cover/land use classification.Thedevelopedspectral/texturalclassificationschema iscomparedwith the classicalapproachusing only spectralinformation.Anaccuracy assessment is carried out which shows that image data with 4IKONOSspectral bands plus NDVI band plus 6 texture bandsachieve an accuracy of 80.01% compared to 63.44% of accuracy achieved by using the few spectral bands only. Furthermore itallows the discrimination of 10 CLC classes.Experimental resultsshow how, starting from available but also binding data (IKONOS imagery andavailable Regional grounddata), a classification schema can be developed with enhanced performance andstrongrelationto the specific setup
