Hydrated minerals (HMs) serve as key indicators of geological processes and the historical presence of water on Mars. Detecting their distribution and composition via remote sensing is therefore crucial for advancing our understanding of Martian aqueous environments. Current detection approaches typically rely on predefined spectral libraries or endmembers extracted directly from images, often overlooking signals from low-abundance components. This leads to low automation in the detection process and unstable inversion results for HM identification. To address these limitations, we propose a novel framework named sequential hydrated mineral detection (SHMD). This method adaptively extracts endmembers from distinct regions using qualitative data without requiring prior information. An automated endmember extraction process further ensures the accuracy and precision of the HM inversion. In this study, experiments were conducted within three regions of the Perseverance rover's candidate landing site. Four categories of samples were labeled in Mawrth Vallis data, and classification accuracy was evaluated using a HM abundance map. Comparative experiments demonstrate that the proposed SHMD outperforms the state-of-the-art methods in terms of detection accuracy, particularly in regions with medium and low abundance of HMs.
SHMD: A Sequential Approach for Martian Hydrated Minerals Detection and Abundance Mapping
Gamba P.
2026-01-01
Abstract
Hydrated minerals (HMs) serve as key indicators of geological processes and the historical presence of water on Mars. Detecting their distribution and composition via remote sensing is therefore crucial for advancing our understanding of Martian aqueous environments. Current detection approaches typically rely on predefined spectral libraries or endmembers extracted directly from images, often overlooking signals from low-abundance components. This leads to low automation in the detection process and unstable inversion results for HM identification. To address these limitations, we propose a novel framework named sequential hydrated mineral detection (SHMD). This method adaptively extracts endmembers from distinct regions using qualitative data without requiring prior information. An automated endmember extraction process further ensures the accuracy and precision of the HM inversion. In this study, experiments were conducted within three regions of the Perseverance rover's candidate landing site. Four categories of samples were labeled in Mawrth Vallis data, and classification accuracy was evaluated using a HM abundance map. Comparative experiments demonstrate that the proposed SHMD outperforms the state-of-the-art methods in terms of detection accuracy, particularly in regions with medium and low abundance of HMs.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
