A Fully-Automated Framework for Mineral Identification on Martian Surface Using Supervised Learning Models

Abstract

The availability of various spectral libraries for CRISM (Compact Reconnaissance Imaging Spectrometer for Mars) data on NASA PDS (Planetary Data System) hugely facilitated the research on the surface mineralogy of Mars, however, building supervised learning models for mineral mapping appears to be challenging due to the lack of ground-truth/training data. In this paper, an automated framework is presented that classifies the spectra in a CRISM hyperspectral image using supervised learning models, where the required training data is produced by augmenting the mineral spectra available in the MICA (Minerals Identified in CRISM Analysis) spectral library, that keeps the key absorption signatures in the mineral spectra intact while providing adequate variability. The framework contains a pre-processing pipeline that in addition to some conventional pre-processing steps includes a new feature extraction method to capture the information of the most distinguishable absorption patterns in the spectra. The proposed framework is validated on a set of CRISM images captured from different locations on the Martian surface by using different types of supervised learning models, like random forests, support vector machines, and neural networks. An uncertainty analysis of the different steps involved in the pre-processing pipeline is provided, as well as a comparison of performances with some of the previously used methods for this purpose, which shows this framework works comparably well with a mean accuracy of around 0.8. Interactive mineral maps are also provided for the detected dominant minerals.

Mineral classification on Martian surface using CRISM hyperspectral data: a survey

“…In the literature, researchers have employed several smoothing techniques to address this issue, including window-median filtering and Savitzky-Golay filtering, with the aim of mitigating such disturbances. 6,67,68 In this section, the two specific challenges associated with the spectra in CRISM hyperspectral images, the issue of mixed spectra and the presence of the continuum resulting from this, are examined in detail, aiming to provide a comprehensive understanding of their implications.…”

“…11, obtained from the Jezero Crater region, has been used in Ref. 68 for mineral identification purposes. Using the proposed ANN model, the dominant minerals detected in the scene include Mg-carbonate, HCP, and Mg-Smectite, which align with findings from a previous study in Ref.…”

“…Despite applying atmospheric and photometric corrections, residual fluctuation noises in the form of spikes or small kinks can persist in spectra, potentially obscuring significant features and introducing errors in scientific analysis. In the literature, researchers have employed several smoothing techniques to address this issue, including window-median filtering and Savitzky–Golay filtering, with the aim of mitigating such disturbances 6 , 67 , 68 . In this section, the two specific challenges associated with the spectra in CRISM hyperspectral images, the issue of mixed spectra and the presence of the continuum resulting from this, are examined in detail, aiming to provide a comprehensive understanding of their implications.…”

“…These labeled pixels can prove to be a valuable asset for tasks, such as nonlinear noise reduction, operating as the training set for mineral identification models, and validated using TRDR or MTRDR data. Another approach 68 involved creating training data by augmenting the spectra available in the MICA spectral library keeping the key absorption signatures in the mineral spectra unaltered while providing sufficient diversity. This allowed for the development of neural network models for mineral identification using CRISM data.…”

Application of Image Sensing System in Mineral/Rock Identification: Sensing Mode and Information Process

“…Mineral/rock identification is an important part of the geological analysis. [ 1 ] The traditional identification method is to manually label samples according to their appearance and compare them with known minerals/rocks to determine their categories. [ 2 ] Although manual identification is simple and feasible, it consumes higher labor costs and has limited accuracy.…”