Imaging spectroscopy for geological mapping

1
Imaging spectroscopy for geological mapping
B. Rivard, D. Rogge, and J. Zhang EOSL,
University of Alberta
2
Outline
Part 1
Mapping in the north south Baffin island
Exploration in the Abitibi gold belt
Part 2
Parallel Developments for underground mining
3
Mapping in the north Project Area
4
Why use Hyperspectral Data
Geological mapping in the arctic is time
intensive and costly, primarily owing to poor
accessibility.
Necessity to pursue effective reconnaissance
mapping methods
Extensive bedrock exposure and limited continuous
vegetation.
5
Generating Geological Maps
Spectral mixing
Spectral mixture analysis (SMA)
Endmember selection
Resulting fractional and geological maps
6
Spectral Mixing
Spectral mixing of image components (within
pixels)
Image pixels are rarely pure spectra.
Can greatly affect classification.
Spectral identification becomes difficult for
mixed pixels.
7
Spectral Mixing
Mixing characteristics impact choice of
classification method
1) Low degree of spectra mixing (e.g. sharp
contacts between components).
Cluster-based spectral classification methods
2) High degree of spectra mixing (e.g. subpixel
mixing common).
Spectral mixture analysis
8
Hyperspectral Endmembers
Spectral library
Acquired under laboratory conditions
Requires knowledge of the area to successfully
select EM. Not all EM components in the project
area may be accounted for.
Field spectra
Accurate representation of area EM.
Requires field work to collect spectra. Not all
EM components in the area may be accounted for.
Image end-members
Collected under the same atmospheric conditions.
Image EM spectra may be a mixed pixel. Requires
relatively pure EM pixels occur within the area.
9
Image Endmember Selection
Spectral-based methods
Iterative error analysis (IEA)
Pixel purity index (PPI)
Integrated spatial-spectral based methods
Spatial-spectral pixel purity index (SSPPI)
10
Image Endmembers using IEA
11
Comparison with Field Spectra
12
Fractional Abundance Maps
13
Fractional Abundance Maps
14
Comparison with Geological Map (1100 000)
15
Future Directions for mapping
Further substantiate merits of integrated
spatial-spectral methods to identify more
lithologies
16
Future Directions for mapping
Provide test of sequential endmember extraction
and purification
Extract the most extreme pixels through
orthogonal projectionendmember candidates Based
on the simplex volume and geometry Automatic and
fast
Beginning with the endmember candidates,
iteratively enlarge the simplex volume to include
all data points
17
Detection of alteration zones in the Abitibi
Mine Bousquet
Gallen Mine
18
Detection of alteration zones

• Endmembers
• Mean spectra of Bousquet alteration
• Mean spectra of felsic lava
• R Alteration spectra
• G Felsic Lava
• B Felsic Lava
• Detection of
• Gallen mine and associated road network
• Two new sites

19
Field checks
CK-1 Old showing, blasted
CK-2 pristine outcrop
20
Spectroscopy of mine waste
Sotiel-Migollas Mine, Sotiel de Coronada, Spain
Processing Plant
Main Tailing site
21
Parallel developments in automated core logging
Long term vision Logging of rock type using
reflectance spectra Geosensing in support of the
automation of routine mine production
22
Wavelet decomposition
Quartz diorite at varying geometry
Effect of grain size and measurement
geometry Non-Compositional variations isolated in
LCP-(b)
23
Isolate suite of features Predict mineral
abundance
Convey in separate images the textural and
compositional properties of samples
34 rock spectra Y axis wavelength X axis SiO2
(78 ? 40)