LULC(land Use Land Cover) Research Based On Multi-source Remote Sensing Data Of Lijiang River Basin

The use of remote sensing data for land use/land cover classification research is one of themost main fields of remote sensing application at present. Now as the development ofmulti-source remote sensing technology and the further improvement of the remote sensingimage resolution, how to combine different features of multi-source remote sensing dataeffectively in the study of land use/land cover classification is a problem of reality.The selection principle of remote sensing data is determined according to the specific taskof the application of the demand for data. In this paper, it depends on the two aspects of thespatial resolution and spectral resolution to select the data source. Using the Landsat7/ETM+multi-spectral data and environmental mitigation satellite hyperspectral (HJ-1A/HSI)data, represented by the Guilin Lijiang River basin karst hoodoo as experimental area, wasstudied by means of surface cover types. There are two aspects studied.（1） In the first place, extract information from the Lijiang River basin as a whole area,which is divided into vegetation area, tree forest, farming area, urban area and water of fivedifferent LULC land use/land cover types by classification and the research. Landsat7ETM+multi-spectral data takes the method of maximum likelihood classification scheme, as well asthe Environment and disaster mitigation satellite hyperspectral (HJ-1A/HSI) data takes themethod of spectral Angle mapping (SAM) classification. Finally, compare the classificationresults of the two methods, analyzed from the two aspects which is image direct comparisonmethod and classification accuracy.The overall classification accuracy of Maximum likelihood of reached80.79%, as well asthe overall precision of SAM method was86.2745%. The maximum likelihood method andSAM method classification of forest land, water, and the urban product accuracy has reached95.09%,99.80%,92.49%and95.09%,86.90%,95.17%respectively, it meets the accuracyrequirement of the large scale terrain surface coverage types to a certain extent.（2）To further the HJ-1A hyperspectral remote sensing data mining application in theclassification of surface coverage value, this article has done further detailed classificationexperiment on the Lijiang River watershed vegetation type rich regions which is interested in. The vegetation in the study area classified as fir forest, birch forest, bamboo forest and pineforest four types by taking the same SAM spectral Angle mapping method. Through theclassification results, it has verify the accuracy of classification result with the localvegetation known data.Research shows that, Landsat ETM+multi-spectrum data should be used when remotesensing image of land use/land cover is divided into five basic types. For the classification ofvegetation type fine when required, HJ-1A hyperspectral data can be of better use.