T146 The G.K. Gilbert Award Session I

Tuesday, 12 October 2021
8:00 AM - 12:00 PM
Portland Ballroom 256 (Hybrid Room)

AUTOMATED IDENTIFICATION OF NOVEL MINERAL SPECTRA IN THE CRISM IMAGE DATABASE USING OPEN SET CLASSIFICATION METHODS

Paper number: T146-11
Program: Technical Sessions
Time: 10:25 AM – 10:40 AM

Authors

Abstract

Recent advances in image processing [4] and hyperspectral mapping [1], coordinated with mineral detections on the surface, are facilitating improved characterization of aqueous and igneous minerals in several regions of Mars. In our prior work, [1] we were able to accurately map over 30 known spectral prototypes identified by the community in the CRISM image database. A related task corresponds to detect novel mineral spectra present in the CRISM images and adding these spectra to the known mineral library. Open set techniques [2][3] are ideally suited to single out spectral signatures that do not belong to known mineral classes. Starting with our current database of detected signatures, we reanalized CRISM images is analyzed with an open-set classifier for the presence of novel spectra which do not belong to any of the known classes. We highlighted several groups of unknown spectra and discovered novel classes by selecting groups that presented novel not previously observed combinations of spectral features. These novel candidate prototypes are then added to the library for mapping and shown to experts for accurate labeling. Using this technique we identified species that exhibited subtle variations over known species in several regions such as Gale crater, Mawrth Valllis, Ius Chasma and others. Examples include jarosite, kaolinite-like spectra mixed to other hydrated phases. A spectrum similar to Monohydrated sulfate was identified , but with significant differences in the 1.2-1.8 µm range and the shape of the band at 1.9 µm. A possible mixture of Poly-hydrated and Mono-Hydrated sulfates was also detected. We also detected a spectrum showsing a doublet with absorptions at 2.22 and 2.28 µm.

REFERENCES

[1] Saranathan, A.M & Parente, M., (2021) Icarus, 355, 114107

[2] Bendale, A & Boult, T.E., (2016), IEEE CVPR.

[3] Shu, L., Xu, H., and Liu, B., (2017), arXiv.

[4]. Saranathan A.M. & M. Parente (2021) Adversarial feature learning for improved mineral mapping of CRISM data, Icarus, 355, 114107

Geological Society of America Abstracts with Programs. Vol. 53, No. 6

doi: 10.1130/abs/2021AM-369845

© Copyright 2021 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.