[An accurate approach to hyperspectral mineral identification based on naive bayesian classification model]

Area of Science:

• Geoscience
• Mineralogy
• Remote Sensing

Context:

• Spectral absorption features of minerals, particularly hydrothermal alteration minerals, often overlap.
• Factors like spectral mixing complicate accurate mineral identification.
• Existing methods struggle with spectral ambiguity, leading to misidentification.

Purpose:

• To develop an accurate hyperspectral mineral identification approach.
• To overcome challenges posed by similar spectral signatures and spectral mixing.
• To improve the reliability of mineral identification in geological surveys.

Summary:

• A naive Bayesian classification model was developed for hyperspectral mineral identification.
• The model analyzes absorption feature position, depth, and continuum slope.
• Performance was evaluated using muscovite and kaolinite, comparing against spectral angle matching, binary encoding, and spectral feature fitting.

Impact:

• The proposed naive Bayesian approach demonstrates superior performance in differentiating minerals with similar spectra.
• It achieves higher classification accuracy compared to traditional methods.
• This enhances the precision of mineral resource exploration and geological mapping.