Research On The Distribution Of Karst Rocky Desertification In Eastern Yunnan Based On Quantitative Analysis Of GF-5 Hyperspectral Images

Karst Rocky Desertification is in the ecological environment of subtropical karst areas.Due to natural evolution and excessive human activities,serious soil erosion and ecological degradation of exposed bedrock are caused,which seriously threaten the regional ecological environment and global ecological balance.Therefore,it is necessary to have a clear understanding of the Rocky Desertification in karst areas.Obtaining continuous spectral information of ground features through hyperspectral remote sensing can advance qualitative analysis to quantitative or semi-quantitative,so as to conduct detailed investigations on Karst Rocky Desertification areas.In order to be able to quantitatively analyze the detailed Karst Rocky Desertification in the study area,the idea of divide and conquer was first introduced,and the big problem of Karst Rocky Desertification was decomposed into the content of lignin in the canopy vegetation,the content of iron oxide,alumina,and the bedrock.Quantitative inversion of calcium carbonate content.Combined with slope data,urban buildings,roads,water bodies,geology and vegetation index and other multiple data,the distribution of Karst Rocky Desertification in the study area was mapped.First,a detailed field survey was conducted in the study area,and 107 ground samples were collected uniformly and coordinate information was recorded.The content of the samples was analyzed by spectrophotometer and X-ray fluorescence spectrometry.Then the GF-5 AHSI hyperspectral image is preprocessed to remove the bad band,radiometric calibration and orthorectification,and then extract the spectral information of the collected points on the spot.The basic principles of Karst Rocky Desertification spectra were deduced,and the double frequency peaks of aromatic cyclic hydrocarbon saturated C-H functional group molecules and alkyl unsaturated C-H functional groups in the canopy vegetation lignin were deduced at 1650 nm and 1746 nm,and the canopy vegetation lignin profile was constructed.The index is used to invert the lignin content of canopy vegetation.Due to the low fitting accuracy of the characteristic wavelengths derived from the theory of soil iron oxide,alumina and calcium carbonate in bedrock,a LASSO regression algorithm using principal component analysis combined with feature selection and machine learning is proposed.The aluminum content and the calcium carbonate content in the bedrock were quantitatively inverted.Predict the lignin content of the canopy vegetation in the study area,the content of iron oxide and alumina in the soil,and the content of calcium carbonate in the bedrock,combined with the multi-source data of urban buildings,roads,DEM,etc.and field surveys of Karst Rocky Desertification,to give evaluation indicators System and superposition analysis are performed to get a detailed status of Karst Rocky Desertification in the study area,and briefly analyze the causes of local Karst Rocky Desertification.Accuracy evaluation of the results after overlay analysis was conducted through397 field survey record points.The overall classification accuracy OA is 0.9091,and the statistical Kappa is 0.9078,which can meet the accuracy requirements.