Research On Key Technologiesof Airborne Imaging Hyperspectral Remote Sensing And Its Application

Airborne imaging hyperspectral remote sensing technology has been applied successfully in mineral exploration,basic geological survey,soil quality investigation and other fields.However,there are many uncertainties and low level of quantification in the application of airborne imaging hyperspectral caused by the coupling processes of different factors,such as solar angle,atmospheric conditions,terrain conditions,the vibration of flight platform,the change of environmental pressure and temperature,mixing of ground objects,light scattering characteristics of sensor and small deviation between optical components.This study takes airborne imaging spectrum measurement system CASI/SASI as a case to carry out key technological researches,including spectral reconstruction,information extraction and application.First of all,according to the main source of error from airborne imaging spectral reconstruction and influence factor analysis,one method of airborne imaging spectral reconstruction has been established based on the absorption depth coefficient,radiative transfer model of atmospheric parameters extracted from imaging spectrum data and Minnaert model.Secondly,based on the reconstructed airborne spectral data,an algorithm study was carried out to analyze the composition of rocks and information extraction of organic matter in soil.According to the mixed pixel problem of airborne spectral data,a deep learning neural network forspectral unmixingis proposed,and the spectral machine learning and mixing experiments are carried out based on the mixed spectral data.The experimental results show that the information extraction method can improve the inversion accuracy from the airborne hyperspectral data efficiently.Finally,the spectral reconstruction and information extraction of iron and organic matter content in black soil have been tested and verified in the large-scale iron ore deposit located in Qilian Mountain,Gansu province and the paddy fields of black land in Sanjiang,Heilongjiang province,respectively.Combined with geological background,geological metallogenic theories and relevant regulations,researches on mineral exploration and quality evaluation on black land have been carried out.Some favorable metallogenic zones have been determined.Besides,the soil organic carbon contents and soil quality in Jiansanjiang area have been evaluated.The study shows that the minerals and soil contain a wealth of spectral information which will be captured by the airborne hyperspectral imager.Analysis of spectral information effectively and precisely are meaningful to the conservation and utilization of natural resources.The paper proposed a new method of spectral reconstruction and the mixed spectrum of deep learning neural network.The method could be applied to mineral exploration and quality evaluation on black land.Related results have been verified in our research and further verification is looking forward to.