Flight altitude and data quality in drone-based hyperspectral imaging for grade control in open-pit mining

Grade control decisions in open-pit mining determine the destination of mined material (e.g. processing, waste). Blast movement modelling translates in situ resource models into muckpile projections that define material units at the point of excavation. However, these projections are rarely checked against independent datasets after blasting. Drone hyperspectral imaging (HSI) can provide such a check by mapping mineralogy directly on the muckpile surface. Yet signal quality is seldom assessed in mining HSI applications, where signal-to-noise ratio (SNR) is generally taken as the nominal sensor value, disregarding both its variation across the spectrum and the effect of acquisition choices on the signal. This study examines whether flight altitude affects the SNR of drone-HSI over a muckpile, evaluated at the absorption features diagnostic of the target mineralogy, to establish whether lower-altitude missions with higher spatial resolutions are justified. HSI data were acquired over a single muckpile at three altitudes (40, 80 and 120 m) under identical sensor settings, with mosaic pixel sizes of 3, 5 and 6 cm respectively. The drone-HSI system comprises a NOA hexacopter (Acecore Technologies) carrying a HySpex Mjolnir VS-620 sensor (Norsk Elektro Optikk). Data were georectified and converted to at-surface reflectance, masked for shadows and oversaturated pixels, and restricted to the ground common to all three datasets. Per-band SNR was then estimated from the image data using identical processing parameters. SNR was highest at 80 m in all diagnostic windows and lowest at 40 m. The 40 m mission required 2.9–3.5 times the imaging time and data volume of the others. Under the illumination of each flight, lower altitude did not improve signal quality, suggesting that missions should be justified radiometrically, not by pixel size. These results provide the data-quality baseline for comparing HSI products against muckpile projections in the next stage of this research.