研究目的
To develop a method for the automated extraction of both mineralogical and structural information from hyperspectral drill-core scans to characterize mineralized bodies, specifically for vein-hosted mineralization in porphyry systems.
研究成果
The proposed methodology integrates mineralogical and structural feature extraction from hyperspectral images, showing good match with visual observations but requires enhancements for thickness variability and automation. Future work will incorporate machine learning for endmember extraction and unmixing.
研究不足
The parameter σ for Gaussian convolution affects detection of veins of different thicknesses; spatial resolution limits detection of fine veins and vein clusters; noise from endmember analysis complicates extraction; improvements needed for automation and handling varied vein sizes.
1:Experimental Design and Method Selection:
A parallel framework is used, combining mineral mapping via linear spectral unmixing and ridge detection for vein extraction.
2:Sample Selection and Data Sources:
Drill cores with porphyry stockwork mineralization are scanned using a hyperspectral sensor.
3:List of Experimental Equipment and Materials:
SisuROCK drill-core scanner, AisaFENIX VNIR-SWIR hyperspectral sensor, ENVI software version
4:Experimental Procedures and Operational Workflow:
Data acquisition involves scanning drill cores; processing includes MNF transform, PPI, endmember extraction with n-D visualizer, SAM classification, linear spectral unmixing to create abundance maps, and ridge detection using Hessian matrix analysis and Steger algorithm.
5:Data Analysis Methods:
Eigenvalue analysis of Hessian matrix, thresholding based on eigenvector magnitudes, and simplification using Visvalingam algorithm.
独家科研数据包,助您复现前沿成果,加速创新突破
获取完整内容