Applications of Ground- and Drone-Based Hyperspectral Remote Sensing in Mining and Metallurgical Environments

Abstract

This research tested hyperspectral remote sensing for large-scale mineral mapping on benches, highwalls, and leach pads in two active mines near Tucson, Arizona. The objective was to determine best practices for applying hyperspectral remote sensing in mining and metallurgical environments. Research topics included testing different procedures for building libraries, different methods of pre-processing, different classification methods, and examining how conditions such as variations in mineralogy and material type, illumination, angle of exposure, and moisture content affect the results. The results show that hyperspectral remote sensing can be a reliable technique for mining companies to map the distribution of spectrally active minerals on a large scale and describes best practices for different scanning areas. The tripod-mounted scan is suitable for steep highwalls and benches, while drone-mounted scans work best for scanning flat areas, like leach heaps. Closer scan and improves identification, particularly of materials with low spectral activity. Different smoothing and binning approaches were used to increase the signal/noise ratio of spectral data. Both of them can be used for pre-processing, but binning is prone to causing misidentification, and smoothing is more reliable. Comparison of the scan results classified by Spectral Angle Mapper (SAM) and Spectral Feature Fitting (SFF) indicates that SAM yields more accurate and comprehensive identifications. All scans show that variations in illumination intensity and moisture content affect mineral identifications, with the results being most serious for drone-based binned spectra classified with a low spectral angle threshold.