Forward Spectral Mixture Modelling for Mineral Mapping Applications

Mineral mapping is crucial for geological and resource management, requiring precise identification of minerals in rocks and ores. Remote sensing techniques, particularly hyperspectral imaging (HSI), play a critical role in analyzing spectral signatures for geological feature mapping. However, multiple minerals within a single HSI pixel complicate mineral exploration due to mixed spectra. This study employs a forward mixture (FM) modelling approach to address spectral unmixing challenges, recovering pure constituent spectra (endmembers) and estimating their proportions (abundances). A synthetic mixed spectral library was created by combining endmember spectra from the United States Geological Survey (USGS) and Johns Hopkins University (JHU) spectral libraries, capturing various mineral interactions and abundance ratios. These spectra were mapped onto AVIRIS-NG hyperspectral image data using Spectral Feature Fitting (SFF) and Spectral Angle Mapper (SAM). Results show that forward spectral mixture modelling significantly improves mineral differentiation and mapping accuracy, achieving scale factors of 0.5 from SFF. Integration with machine learning further enhances exploration capabilities. This advancement supports sustainable resource management and broadens the applications of mineral mapping