Research On The Method Of Martian Surface Hydrous Minerals Detection With Crism Data

The research on the formation and distribution of hydrous minerals on Mars can provide basis for the formation of Mars crust,heat loss process,magmatism,hydro-thermal action and sedimentation.Therefore,the identification of hydrous minerals on the surface of Mars is crucial to the prediction of Mars' past water environment and the extraterrestrial life searching.With the development of Mars exploration technology,abundant highresolution and high-precision Mars image data provide necessary data support for the identification of hydrous minerals on Mars.As the lack of Mars field mineral samples,this paper study different approaches on the identification of hydrous minerals by Mars hyperspectral image data.In this paper,CRISM(Compact Reconnaissance Imaging Spectrometer for Mars)data is employed to identify and classify hydrous minerals in Jezero and Holden craters on Mars.Two different methods are developed to identify hydrous minerals on Mars,their formation mechanism and spectral characteristics are discussed,and the early paleowater environment on Mars are predicted.For the mineral identification,the traditional extraction method of spectral information of earth mineral characteristics is adopted to identify minerals by comparing with the Martian mineral spectral database established by NASA,details include(1)noise reduction through MNF(Minimum Noise Fraction)transformation,pure end elements are extracted by PPI(Pixel Purity Index),and the identification of Mars mineral spectral database is carried out by ROI(region of interest)mapping.(2)Target minerals are identified by SAM(Spectral Angle Mapping)based on the spectral curves of each target end element and the threshold transformation of the spectral angle in the process of the spectral curves of the target minerals in the spectral library.The advantage of this method is that it is easy to operate and mature in the process of mineral identification on Earth.However,it is rarely used on Mars at present.It can identify hydrous minerals on Mars more accurately by referring to the Mars Mineral Spectrum Library established by NASA.On the other hand,Mars mineral information is abtained utilizing Earth ground truth and computer vision technology in deep learning which has been rapidly developped in recent years.Mars hydrous minerals identification based on convolutional neural network is achieved by well-designed network structure combined with domain knowledge.The final OA and KAPPA coefficients show that the recognition result of Mars minerals based on convolutional neural network can reach more than 80%.Furthermore,the Mars hydrous minerals identification method based on convolutional neural network are compared with those of traditional feature spectrum information extraction methods,the identified Mars hydrous minerals distribution area are highly consistent.