Predicting spectral channels related to mineralogy from CRISM spectral bands using deep learning

Aman SAHANI

Master's thesis

Abstract:

The Mars Reconnaissance Orbiter (MRO) is equipped with a suite of instruments that collectively generate a vast and diverse array of imaging data for comprehensive investigations of the Martian surface. Notably, the hyper-spectral Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) captures reflective information across multiple spectral bands. However, CRISM coverage is partial, limiting the availability of full spectral and spatial resolution data and the instrument is no longer working. In contrast, the Colour and Stereo Surface Imaging System (CaSSIS), characterized by high spatial resolution and diverse color capabilities, covers larger areas of the Martian surface. The goal of this project is to improve the accuracy of predicting specific spectral signatures, such as carbonates, using spectral reflectance data. I want to strip CRISM images to CaSSIS bands and then use the generative abilities of diffusion model to predict the carbonate signature.

Abstract:

The Mars Reconnaissance Orbiter (MRO) is equipped with a suite of instruments that collectively generate a vast and diverse array of imaging data for comprehensive investigations of the Martian surface. Notably, the hyper-spectral Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) captures reflective information across multiple spectral bands. However, CRISM coverage is partial, limiting the availability of full spectral and spatial resolution data and the instrument is no longer working. In contrast, the Colour and Stereo Surface Imaging System (CaSSIS), characterized by high spatial resolution and diverse color capabilities, covers larger areas of the Martian surface. The goal of this project is to improve the accuracy of predicting specific spectral signatures, such as carbonates, using spectral reflectance data. I want to strip CRISM images to CaSSIS bands and then use the generative abilities of diffusion model to predict the carbonate signature.

Keywords

Stable Diffusion, Unet, Encoder, Mars Reconnaissance Orbiter, Compact Reconnaissance Imaging Spectrometer for Mars (CRISM), Colour and Stereo Surface Imaging System (CaSSIS), PCA, Carbonates

Jazyk práce: English

Datum vytvoření / odevzdání či podání práce: 20. 8. 2024

Obhajoba závěrečné práce

• Vedúci: Dr. Benedikt Elser

UNIVERSITY OF SOUTH BOHEMIA IN ČESKÉ BUDĚJOVICE

Faculty of Science

Master programme / odbor:
Artificial Intelligence and Data Science / Artificial Intelligence and Data Science