The Geological Exploration Based On The Hyperspectral Linear Mixing Model

SWIR hyperspectral minerals test and analysis technology was a new way of mineral exploration introduced from abroad in recent years.It was a method of hyperspectral remote sensing technology based on various electromagnetic wave absorption characteristics of lithology and mineral assemblage in SWIR. The JPL laboratory had test various types of monomineral and gathered a spectral library.However,the rocks is not with single component,so the spectral response was not from single mineral.We don’t exact know how the mineral assemblage response the SWIR spectral.Thus,the problem of spectral mixing should be sovled in mineral exploration.The author measured the sample from drilling and rocks in Jiama copper polymetallic deposit with hyperspectral short-wave infrared technology.The research focus on the problem of data noise reduction and spectral mixing.The practice study proved that hyperspectral short-wave infrared technology possess great significance in JiaMa geological prospecting and the ore-forming environment inversion.In this paper, the main work and innovation points are as follows:1.Introduce and review the strengths and weaknesses of some classical onedimensional signal noise reduction algorithms systematically. According to the characteristics of the spectral signal noise, combined the empirical mode decomposition(EMD) with wavelet threshold method and raised a self-adapting spectrum EMD threshold denoising algorithm; Improved the conventional noise reduction effect evaluation method,the on this basis of introducing spectral complexity and spectral similarity indexes to the evaluation process,a new realistic evaluation index was built.It was proved that the EMD threshold denoising algorithm posses good and stability effect of noise reduction.2.Induces and summarizes the existing three kinds of mixed pixels decomposition(endmember extraction) algorithms.Put forward a endmember filter algorithm based on fully-constrained least squares.The simulation experiments were done to test the extraction accuracy and the variation of fitting error on the condition of data noise and characteristics weaking.The experimental results showed that the algorithm above has a certain ability to resist noise and relatively sensitive to spectral general reflectance.Thus, the Spectra data normalization should be done to improve unmixing accuracy.3. Hyperspectral data of Jiama drill cores and rocks had been analyzed systematically and distribution model of spectral features had been built on the form of two dimension and three dimension.The geologic setting and chemical analysis results were involved to the model,all shows that the metallogenic conditions were medium-high temperature. The heat source center located near 24-32 lines. The representative minerals are are dispersive biotite, a small amount of Al-poor muscovite, a large amount of magnesium chlorite and gypsum at the bottom of some drills.Peripheral hyperspectral rock exploration model is established. The rocks analysis results shows that there is a concealed granitic body of the depth in XiaPu. According to the result of rock spectral analysis,prospective areas can then be delineated.