1520 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
signatures that can be observed in the electromagnetic spec-
trum. Environmental constraints affect the spectral features
that are observable or detectable. Climate and atmosphere
are part of the radiometric correction/compensation pro-
cess and residual artifacts can be incorporated into the
hyperspectral reflectance data outside of laboratory data
and impact hyperspectral data processing and interpreta-
tion. Vegetation can cover materials in a mining environ-
ment and this coverage can partially or fully obscure a
materials’ spectroscopic signature. Similarly, soil cover and/
or dust affects the observable features of the materials of
interest. In the case of dust, the thickness of dust and the
dust’s mineralogy directly influence the observable and/or
detectable features which can make the material’s spectro-
scopic signatures undetectable.
DATA SELECTION
Geometallurgical objectives and models along with observ-
able and detectable phenomena were discussed in the previ-
ous sections, data selection is an important piece of applying
hyperspectral imaging correctly for achieving the geomet-
allurgical objectives. Hyperspectral data selection can be
broken down in three aspects: phenomenology, collection
platform, and data constraints. Sensor attributes such as
number of channels, full width half max (FWHM) of the
channels, signal to noise ratio (SNR), radiometric quality,
non-uniformity correction, cross track quality, blinking,
bad pixels and other items has been left off because this is
beyond the scope of this paper.
Phenomenology is a material’s spectroscopic signa-
tures that you are trying to measure and should be the pri-
mary criteria in selecting the portion of electromagnetic
spectrum to measure. For example, if you need to measure
clay minerals you may choose the SWIR, or if you are con-
cerned about low releases of methane (CH4) gas as part
of your ESG compliance the LWIR may be your choice
(Zimmerman and Kerekes 2023). Table 1 and Table 2 pro-
vides a starting point for gangue and ore minerals that may
be part of your geometallurgical objectives.
Collection platform selection is a critical part of the
overall data selection process as there are hyperspectral sen-
sors that collect data across a variety of scales (Table 4).
Currently, there are 3 spaceborne hyperspectral sensors (i.e.,
PRISMA, ENMAP, and EMIT). Manned aircraft including
helicopters can carry hyperspectral sensors. Sensor manu-
facturers have started producing smaller hyperspectral sen-
sors capable of flying on Unmanned Aerial Systems (UAS)/
Vehicles (UAV) that are under the 55 lb/25 kg weight limit
authorized by the FAA. Many of the hyperspectral sensors
have built their sensors to work rotary stages mounted on
tripods for ground-based scanning. Laboratory and core
logging setups are similar in that they use a scanning table or
tables for measurements (University of Arizona MIINERS
laboratory, Figure 5). Some of the core scanning systems
are mobile while some of them have longwave (LWIR) or
midwave infrared (MWIR) sensors.
Scale and resolution are linked (Table 4) and important
considerations in answering the geometallurgical objective
and determining if the results can contribute to increas-
ing the information for 3-dimensional modeling. Butcher,
Dehaine et al. (2023) discuss scale and cost per sample
for different analytical techniques including hyperspectral
imagery that can be applied in geometallurgy. Our view of
scale and resolution is different than Butcher’s astsumptions
Table 3. Spectroscopic response definition for minerals and wavelength range (Coulter, Zhou et al. 2017)
Uncertain Available spectral data insufficient
None Non-Diagnostic responses or no responsiveness of mineral in region
Weak Weak or selective response but mineral characterization difficult
Moderate Moderate but better mineral responses in other regions
Good Good response but mixtures can influence mineral characterization
Clear Most suitable region for mineral identification
Table 4. Scale of Imaging Spectroscopy, Map Scale =Raster resolution (in meters (X)) *2 *1000 (Tobler 1987, Tobler 1988)
signatures that can be observed in the electromagnetic spec-
trum. Environmental constraints affect the spectral features
that are observable or detectable. Climate and atmosphere
are part of the radiometric correction/compensation pro-
cess and residual artifacts can be incorporated into the
hyperspectral reflectance data outside of laboratory data
and impact hyperspectral data processing and interpreta-
tion. Vegetation can cover materials in a mining environ-
ment and this coverage can partially or fully obscure a
materials’ spectroscopic signature. Similarly, soil cover and/
or dust affects the observable features of the materials of
interest. In the case of dust, the thickness of dust and the
dust’s mineralogy directly influence the observable and/or
detectable features which can make the material’s spectro-
scopic signatures undetectable.
DATA SELECTION
Geometallurgical objectives and models along with observ-
able and detectable phenomena were discussed in the previ-
ous sections, data selection is an important piece of applying
hyperspectral imaging correctly for achieving the geomet-
allurgical objectives. Hyperspectral data selection can be
broken down in three aspects: phenomenology, collection
platform, and data constraints. Sensor attributes such as
number of channels, full width half max (FWHM) of the
channels, signal to noise ratio (SNR), radiometric quality,
non-uniformity correction, cross track quality, blinking,
bad pixels and other items has been left off because this is
beyond the scope of this paper.
Phenomenology is a material’s spectroscopic signa-
tures that you are trying to measure and should be the pri-
mary criteria in selecting the portion of electromagnetic
spectrum to measure. For example, if you need to measure
clay minerals you may choose the SWIR, or if you are con-
cerned about low releases of methane (CH4) gas as part
of your ESG compliance the LWIR may be your choice
(Zimmerman and Kerekes 2023). Table 1 and Table 2 pro-
vides a starting point for gangue and ore minerals that may
be part of your geometallurgical objectives.
Collection platform selection is a critical part of the
overall data selection process as there are hyperspectral sen-
sors that collect data across a variety of scales (Table 4).
Currently, there are 3 spaceborne hyperspectral sensors (i.e.,
PRISMA, ENMAP, and EMIT). Manned aircraft including
helicopters can carry hyperspectral sensors. Sensor manu-
facturers have started producing smaller hyperspectral sen-
sors capable of flying on Unmanned Aerial Systems (UAS)/
Vehicles (UAV) that are under the 55 lb/25 kg weight limit
authorized by the FAA. Many of the hyperspectral sensors
have built their sensors to work rotary stages mounted on
tripods for ground-based scanning. Laboratory and core
logging setups are similar in that they use a scanning table or
tables for measurements (University of Arizona MIINERS
laboratory, Figure 5). Some of the core scanning systems
are mobile while some of them have longwave (LWIR) or
midwave infrared (MWIR) sensors.
Scale and resolution are linked (Table 4) and important
considerations in answering the geometallurgical objective
and determining if the results can contribute to increas-
ing the information for 3-dimensional modeling. Butcher,
Dehaine et al. (2023) discuss scale and cost per sample
for different analytical techniques including hyperspectral
imagery that can be applied in geometallurgy. Our view of
scale and resolution is different than Butcher’s astsumptions
Table 3. Spectroscopic response definition for minerals and wavelength range (Coulter, Zhou et al. 2017)
Uncertain Available spectral data insufficient
None Non-Diagnostic responses or no responsiveness of mineral in region
Weak Weak or selective response but mineral characterization difficult
Moderate Moderate but better mineral responses in other regions
Good Good response but mixtures can influence mineral characterization
Clear Most suitable region for mineral identification
Table 4. Scale of Imaging Spectroscopy, Map Scale =Raster resolution (in meters (X)) *2 *1000 (Tobler 1987, Tobler 1988)