Nonnegative CP decomposition of multiangle hyperspectral data: a case study on CRISM observations of martian icy surface

Miguel Angel Veganzones (1, 2) , Sylvain Douté (3) , Jérémy E Cohen (2) , Rodrigo Cabral Farias (2) , Jocelyn Chanussot (1) , Pierre Comon (2)

1 GIPSA-SIGMAPHY - GIPSA - Signal Images Physique
2 GIPSA-CICS - GIPSA - Communication Information and Complex Systems
3 IPAG - Institut de Planétologie et d'Astrophysique de Grenoble

The Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) sensor aboard the Mars Reconnaissance Orbiter takes hyperspectral multi-angle acquisitions of Martian surface from the top of the atmosphere (TOA) on visible and infrared wavelengths. The Multiangle Approach for Retrieval of Surface Reflectance from CRISM Observations (MARS-ReCO) defined an innovative TOA radiance model and inversion scheme aimed at correcting for aerosols effects taking advantage of the near-simultaneous multi-angle CRISM observations. Here, we aim to provide validation evidence of MARS-ReCO by unmixing the estimated multi-angle bidi-rectional reflectance (BRF) from highly reflective and anisotropic icy surfaces at high latitudes with grazing illumination, using a non-negative CP decomposition. Obtained results are in accordance with other complementary studies, which compose a collaboration effort to validate MARS-ReCO through the cross-validation of different techniques in the absence of ground truth.

Mots clés: Canonical-Polyadic; nonnegative tensor decomposition; multi-angle hyperspectral; MARS-ReCO; CRISM mission

Domaines: Traitement du signal et de l'image [eess.SP]

Miguel Angel Veganzones, Sylvain Douté, Jérémy E Cohen, Rodrigo Cabral Farias, Jocelyn Chanussot, et al.. Nonnegative CP decomposition of multiangle hyperspectral data: a case study on CRISM observations of martian icy surface. WHISPERS 2016 - 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, Aug 2016, Los Angeles, CA, United States. ⟨hal-01382360⟩