Spectral unmixing for mineral identification in pancam images of soils in Gusev crater, Mars

摘要: Abstract The objective of this work is to propose an automated unmixing technique for the analysis 11-channel Mars Exploration Rover Panoramic Camera (MER/Pancam) spectra. Our approach provide a screening tool identifying unique/distinct reflectance We demonstrate utility in study mineralogy bright salty soils exposed by rover wheels images Gusev crater regions known as Paso Robles (Sols 400,426), Arad (Sol 721), and Tyrone 790). algorithm based on novel derivation Nonnegative Matrix Factorization includes added features that preclude adverse effects low abundance materials would otherwise skew unmixing. In order create full spectrum out left right eye stereo pairs, we also developed new registration procedure rectification disparity calculation images. identified two classes endmember spectra imaged Sols 426 790. One these observed 400 721 has unique spectral shape distinct from most iron oxide, sulfate silicate differs typical martian surface Instead, its character resembles ferric minerals fibroferrite (Fe3+(SO4)(OH) · 5H2O) ferricopiapite ( Fe 3 + ,Al,Mg ) 5 SO 4 6 OH 2 · 20 H O phosphate mineral ferristrunzite Mn , PO . other class less consistent with specific likely mixture altered volcanic material some salts. Further analyses data 790 using anomaly detection detecting additionally suggests identification paracoquimbite (Fe2(SO4)3 · 9H2O). This Pancam S- P-enriched identifies are properties here 0.4–1 μm range.