A UNet Model for Accelerated Preprocessing of CRISM Hyperspectral Data for Mineral Identification on Mars

2025-05-04

AI Summary

Preprocessing of Mars CRISM MTRDR hyperspectral data is computationally intensive and time-consuming (≈1.5 hours per scene), severely limiting mineral identification efficiency and geological interpretation. Method: This paper proposes an end-to-end preprocessing framework based on a UNet autoencoder—the first application of UNet to automated smoothing and continuum removal for Martian hyperspectral data—designed to preserve diagnostic mineral absorption features while drastically accelerating processing. The method integrates spectral augmentation using the MICA library and realistic MTRDR-condition simulation to ensure fidelity and robustness, and incorporates the MICAnet mineral classifier with GPU (NVIDIA T1600) acceleration. Results: Processing time for an 800×800 scene is reduced to 5 minutes—a 18× speedup—while achieving state-of-the-art mineral identification accuracy on CRISM TRDR ground-truth annotated data.

Technology Category

• Machine Learning: Dimensionality Reduction/Feature Selection
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Computer Vision: Remote Sensing / Geospatial AI

Application Category

• Web Mining and Content Analysis: Web data integration and cleaning
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Accurate mineral identification on the Martian surface is critical for understanding the planet's geological history. This paper presents a UNet-based autoencoder model for efficient spectral preprocessing of CRISM MTRDR hyperspectral data, addressing the limitations of traditional methods that are computationally intensive and time-consuming. The proposed model automates key preprocessing steps, such as smoothing and continuum removal, while preserving essential mineral absorption features. Trained on augmented spectra from the MICA spectral library, the model introduces realistic variability to simulate MTRDR data conditions. By integrating this framework, preprocessing time for an 800x800 MTRDR scene is reduced from 1.5 hours to just 5 minutes on an NVIDIA T1600 GPU. The preprocessed spectra are subsequently classified using MICAnet, a deep learning model for Martian mineral identification. Evaluation on labeled CRISM TRDR data demonstrates that the proposed approach achieves competitive accuracy while significantly enhancing preprocessing efficiency. This work highlights the potential of the UNet-based preprocessing framework to improve the speed and reliability of mineral mapping on Mars.

Problem

• Accelerating CRISM hyperspectral data preprocessing for Mars mineral identification
• Reducing computational intensity of traditional spectral preprocessing methods
• Preserving mineral absorption features while automating preprocessing steps

Innovation

• UNet-based autoencoder for CRISM data preprocessing
• Automates smoothing and continuum removal steps
• Reduces preprocessing time from hours to minutes