Study On Extracting Lithological Information From Remote Sensing Data In Saishiteng Area,Qinghai Province

Remote sensing lithology interpretation is to use the spectral characteristics of the feature to highlight different lithologies on the image through the method of lithology information enhancement.It combines with field verification and finally identifies different lithologic units.At present,as remote sensing spatial exploration technology continues to develop and the techniques and methods of lithology identification are also being enriched,single data can identify the collaborative interpretation of multi-source data.Qualitative interpretation by manual visual inspection of the collection of rock spectral information can be quantitatively interpreted by supervised or unsupervised classification methods.The Saishiteng area in Qinghai is part of the tectonic zone on the northern margin of the Caihai,with a concentration of bedrock outcrops.In recent years,studies on the extraction of mineralized alteration information have been carried out mainly in the Yuqia-Dachaidan-Delingha area;there are few studies on the extraction of remote sensing lithologic information in the Saishiteng area,especially for the metamorphic rocks exposed in the Saishiteng area,and no relevant remote sensing geological interpretation work has been carried out.This study combined ASTER,Landsat-8 OLI,Sentinel-2A,and SPOT-7 data,using the Saishiteng area in Qinghai as the research area.The ASTER shortwave infrared band was processed in collaboration with the Landsat-8 OLI,and Sentinel-2A visible near-infrared bands to form the Landsat-8 OLI+ASTER(LA)data and Sentinel-2A+ASTER(SA)data.This study analyzed the standard spectral information of the resampled rocks,formulated the formulae for different rock bands,selected the threshold range of different rock types based on the multiple fractal theory,and obtained the distribution of the main lithologies.In addition,it used the best band index to determine the band combination of each image to highlight the boundaries of different rocks and made use of the principal component analysis to further enhance the identification of different lithologies.By comparing and analyzing the lithologic classification accuracy of three typical machine learning methods,such as support vector machine,maximum likelihood method and random forest,in the alpine valley area,and combining with the field survey,this paper summarized and concluded the automatic lithologic classification technology method in the alpine valley area.The main conclusions and results are shown as follows:(1)The lithology information enhancement method was used to compare ASTER,Landsat-8 OLI and Sentinel-2A data.Due to the advantages of Landsat-8 OLI and Sentinel-2A data in visible-near-infrared bands,the lithology identification ability of Landsat-8 OLI and Sentinel-2A data was superior to ASTER data in alpine and canyon areas,such as the Saishiteng area.(2)GS image fusion algorithm solves the problem of inconsistent spatial resolution of multi-source remote sensing data,improves the spatial resolution of multi-source remote sensing data,and improves the ability of multi-source remote sensing data to extract fine rock mass inside the formation.(3)By co-processing ASTER data with Landsat-8 OLI and Sentinel-2A data,the defects of Landsat-8 OLI and Sentinel-2A data in short-wave infrared band are made up.The ability of Landsat-8 OLI and Sentinel-2A data to identify metamorphic rocks is improved.(4)By comparing the different processing methods of Landsat-8 OLI and Sentinel-2A data,the classification accuracy of different monitoring algorithms was determined to determine the data processing methods,thus improving the efficiency of lithology information identification.