Research Of Lithology Recognition Method With Thermal Infrared Hyperspectral (TASI) Data

The thesis with "The Research and demonstration of thermal infrared hyperspectral mineralization and alteration mineral extraction" project,carry out lithology identification method research for thermal infrared hyperspectral(TASI)data.Use three methods(spectrum matching,lssvm,sparse representation)to identify lithology with the thermal infrared hyperspectral data in Gansu Liuyuan huaheitan and comparative evaluation the results of the three methods.The main contents and results are as follows:(1)A lithology identification method: SDEM be brought up with higher discriminative of similar spectrum.spectral matching algorithm.Using Matlab programming platform for SAM,SID,SDEM algorithm of three kinds of spectral matching method on the implementation(2)Building a least squares support vector machine lithology identification model for the high thermal infrared hyperspectral(TASI)data:Use "one-to-many(1vr)" and "error correction input(ECOC)" multi-class coding method to encode each lithologic category,according to the encoding method to determine the required number of recognizer,set the gaussian radial basis function(RBF)as thekernel function of recognizer and use the network search method on cross validation and the training sample to optimize kernel parameters,in the end get the final lithology recognition model.Based on Matlab programming platform to achieve the above build process,and apply it to the lithology recognition in the study area.(3)proposed a method based on lithology identification sparse representation:add the fisher criterion to dictionary learning of training samples,to build higher separability of dictionaries.After obtaining the learning dictionary,using ordinary sparse coding(SRC)and kerner sparse coding(KSRC)to sparse representative of rock spectrum and indicated that the lithology based on sparse residual.Based on Matlab programming platform to achieve the sparse representation lithology identification process,and apply it to the lithology recognition in the study area.(4)Comparative evaluation these three kinds of methods with the recognition effect,efficiency and robustness,Selecting LSSVM(1vr)and KSRC as the optimal lithologic identification methods for thermal infrared hyperspectral(TASI)data,but also get the lithology recognition results in the study area,find out the distribution of lithology in this region.