|
|
|
EOM June 2005 > Features
Evaluation of Hyperspectral Remote Sensing
as a Means of Exploration Geology
M. Rajesh Kumar and T.N. Singh
Because of the unique spectral characteristics of many alteration and rock-forming minerals, hyperspectral remote sensing can make a significant contribution to the field of exploration geology. Classical geologic mapping and mineral exploration utilize physical characteristics of rocks and soils such as mineralogy, weathering characteristics, geochemical signatures, and landforms to determine the nature and distribution of geologic units and to determine exploration targets of metals and industrial minerals. Subtle mineralogical differences—often important for distinguishing between rock formations or between barren ground and potential economic ore—are often difficult to map in the field. Hyperspectral remote sensing, the measurement of the Earth's surface in up to hundreds of spectral images, provides a unique means of remotely mapping mineralogy. A wide variety of hyperspectral data are now available, along with operational methods for qualitatively analyzing the data and producing mineral maps. This article illustrates the potential of these data and how they can be used as a tool to aid geological mapping and exploration.
Introduction
Imaging spectrometry data or hyperspectral imagery (HSI) measure the reflectance or emissivity of the Earth's surface in many spectral bands, providing both spatial images and contiguous spectral coverage over selected spectral ranges. Hyperspectral data have been used for detailed mapping of materials for more than 15 years for geology. The solar spectral range, 0.4 to 2.5 micrometers, provides abundant information about many important Earth-surface minerals and research has proven the ability of hyperspectral systems to uniquely identify and map many of these minerals, even in sub-pixel amounts. Research has shown that imaging spectrometry principles and mapping capabilities are extensible to many other disciplines.
Imaging spectrometers acquire images in up to several hundred contiguous spectral bands simultaneously. Such data are well-suited to the identification of target materials having spectral reflectances distinguished primarily by narrow spectral features and are particularly important in discriminating between materials for which these features are limited to the same small region of the spectrum. One application requiring such data is the mapping of clay minerals. The reflectance spectra of these minerals have narrow absorption features (down to 10 nm in width) in the 2050 nm to 2350 nm range. In addition, the exact wavelength location and shape of a particular absorption feature may vary for an individual mineral type depending on the chemical composition of the sample in question. Such information can be extracted from imaging spectrometer data by an expert analyst.
There are an increasing number of airborne imaging spectrometers in operation—including the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), covering the spectral range between 400 nm and 2500 nm with 224 bands; Probe-1 with 128 bands over the range from 440 nm to 2543 nm; the Compact Airborne Spectrographic Imager (CASI), with up to 288 bands in the visible and near infrared part of the spectrum (400 nm to 1000 nm); and the Short wave infrared Full Spectrum Imager (SFSI), with 235 bands covering the spectral range from 1200 nm to 2450 nm.
One of the methods commonly used for the analysis of imaging spectrometer data is spectral unmixing. In general, a pixel observed by the remote sensing instrument consists of mixed materials. The hypothesis underlying linear unmixing is that the spectral radiance measured by the sensor consists of the radiances reflected by all of these materials, summed in proportion to the subpixel area covered by each material. To the degree that this hypothesis is valid, and that the "endmembers" are given by the reference spectra of each of the individual pure materials, and under the condition that these spectra are linearly independent, then in theory one can deduce the makeup of the target pixel by calculating the particular combination of the endmember spectra required to synthesize the target pixel spectrum. In practice there are a number of factors which act to confuse the issue. Some of these are sensor artifacts, sensor noise, atmospheric effects, solar incidence and terrain slope angles, surface roughness, and other radiometric influences.
|
One possible source for the endmember spectra are libraries of spectral reflectances. The risk in using such library spectra in the unmixing operation is that the library spectra are rarely, if ever, acquired under the same conditions as the airborne data. The size of the particles constituting the mineral sample and the illumination conditions are but two of the variables that can have a significant effect on the resulting spectra. A better match will be obtained if the endmember spectra are taken from the image cube under analysis.
Mineral Identification and Mapping
Imaging spectrometry is being used increasingly for mapping surface alterations on the Earth's surface. This is because numerous recent results have shown that spectral analysis of both field and airborne imaging spectrometer data can provide useful mineralogical and geochemical information for geological mapping and exploration. In addition, the image processing techniques developed to facilitate surface compositional mapping using data compression and spectral unmixing techniques are now more robust.
Alteration minerals, e.g. clays, are of particular interest to the exploration and mining industry because of their association with occurrences of precious metals such as gold or silver. To map these alteration minerals, many researchers have developed methodologies and software which utilize a technique called spectral unmixing to determine the composition of scene pixels consisting of a mixture of different materials.
Figure 1: AVIRIS images of the Cuprite mining district in southern Nevada. Image courtesy of NASA. Click on image to see enlarged.
Analysis of Hyperspectral Data
Spectral Angle Mapper Classification
The spectral angle mapper (SAM) classification is an automated method for directly comparing image spectra to known spectra (usually determined in a lab or in the field with a spectrometer) or to an endmember. This method treats both the questioned and known spectra as vectors and calculates the spectral angle between them. This method is insensitive to illumination since the SAM algorithm uses only the vector direction and not the vector length. The result of the SAM classification is an image showing the best match at each pixel. This method is typically used as a first cut for determining the mineralogy and works well in homogeneous regions.
SAM is different from standard method classifications because it compares each pixel in the image with every endmember for each class and assigns a ponderation value between 0 (low resemblance) and 1 (high resemblance). Endmembers can be taken directly from the images or from signatures measured directly in the field or laboratory. The main advantage of the SAM algorithm is that it is an easy and rapid method for mapping the spectral similarity of image spectra to reference spectra. It is also a powerful classification method because it represents the influences of shading effects to accentuate the target reflectance characteristics. The main disadvantage of this technique is that it does not consider the sub-pixel value. The spectral mixture problem can become problematic because most of the Earth's surface is heterogeneous.
Spectral Unmixing/Matched Filtering
Most surfaces on the Earth, geologic or vegetated, are not homogeneous, which results in a mixture of signatures characterized by a single pixel. The mathematical model used to determine the relative abundance of materials depends on how theymix on the surface. If the spectral mixing is significantly larger than the mixing of the signatures, it can be represented as a linear model. However, if the mixing is microscopic, then the mixing models become more complex and non-linear.
The first step to determining the abundances of materials is to select endmembers, which is the most difficult step in the unmixing process. The ideal case would consist of a spectral library, which consists of endmembers that, when linearly combined, can form all observed spectra. A simple vector-matrix multiplication between the inverse library matrix and an observed mixed spectrum gives an estimate of the abundance of the library endmembers for the unknown spectrum.
N-Dimensional visualization techniques can be used to select endmembers within a scene. Figure 2 is a 2-dimensional representation of endmember selection. Extreme pixels which ultimately correspond to endmembers can be determined by rotating this scatter plot in n-dimensions.
Figure 2: 2-Dimensional scatter plot of Eigenvectors 1 & 2. Image courtesy of USGS. Click on image to see enlarged.
Matched filtering is based on a well-known signal processing method and creates a quick means of detecting specific minerals based on matches to specific library or endmember spectra. The matched filtering algorithm maximizes the response of a known endmember while suppressing the response of the background. The resultant matched filtering resembles the results from the linear unmixing methods and are usually represented as a grayscale image with values ranging from 0 to 1, which corresponds to the relative degree of the match.
Other Classification Techniques
Classification and feature extraction methods have been commonly used for many years to map minerals and vegetative cover in multispectral data sets. Conventional classification methods, such as a Gaussian Maximum Likelihood algorithm, cannot be applied to hyperspectral data due to the high dimensionality of the data.
The difficulty in using many classification methods based upon conventional multivariate statistical approaches is that many of these methods rely on having a non-singular class and specific covariance matrices for all classes. When working with high-dimensional data sets, it is likely that the covariance matrices will be singular when using a limited (with respect to the number of input bands) number of training samples.
A nonparametric classifier, such a neural network, and other feature extraction methods can be used to accurately classify hyperspectral images. Feature extraction methods, such as decision boundary feature extraction (DBFE), can extract the features necessary to achieve classification accuracy while reducing the amount of data analyzed in feature space. 
Sources
J.B. Adams, M.O. Smith, and P.E. Johnson, "Spectral Mixture Modeling: A New Analysis of Rock and Soil Types at the Viking Lander Site," Journal of Geophysical Research, Vol. 91, pp. 8098-8112, 1986.
C.D. Anger, S. Achal, T. Ivanco, S. Mah, R. Price, and J. Busler, "Extended Operational Capabilities of casi", Proceedings of the Second International Airborne Remote Sensing Conference, San Francisco, California, pp. 124-133, 1996.
J. A. Benediktsson, , J. R. Sveinsson, and K. Arnason, "1995 Classification and feature extraction of AVIRIS data", IEEE Trans. Geoscience and Remote Sensing, 33(5), pp. 1194-1205.
J. W. Boardman and F. A. Kruse, , 1994, "Automated spectral analysis: A geologic example using AVIRIS data, north Grapevine Mountains, Nevada," in Proceedings, Tenth Thematic Conference on Geologic Remote Sensing, ERIM, Ann Arbor, MI, p. I-407 - I-418
J.W. Boardman, "Analysis, Understanding and Visualization of Hyperspectral Data as Convex Sets in n-Space." In Proceedings of the International SPIE Symposium on Imaging Spectrometry, SPIE Vol. 2480, Orlando, Florida, pp.23-36, 1995.
R. N. Clark, , T. V. V. King, , M. Klejwa, and G. A Swayze, , 1990, "High spectral resolution spectroscopy of minerals," Journal of Geophysical Research, v. 95, no. B8, p. 12653-12680.
O.A. De Carvalho, and P.R. Meneses (2000) Spectral Correlation Mapper (SCM); an improvement on the Spectral Angle Mapper (SAM), summaries on the 9th JPL Airborne Earth Science workshop, JPL publication 00-18, 9p.
A. F. H. Goetz, , G. Vane, J. E. Solomon, and B. N. Rock, 1985, "Imaging spectrometry for earth remote sensing," Science, v. 228, p. 1147 - 1153
F.A. Kruse, A.B. and Lefkoff, 1993, "Knowledge-based geologic mapping with imaging spectrometers," Remote Sensing Reviews, Special Issue on NASA Innovative Research Program (IRP) results, v. 8, p. 3 - 28.
F.A. Kruse, J.W. Boardman, and J.F. Huntington, 1999, "Fifteen Years of Hyperspectral Data: northern Grapevine Mountains, Nevada," in Proceedings of the 8th JPL Airborne Earth Science Workshop: Jet Propulsion Laboratory Publication, JPL Publication 99-17, p. 247 - 258
F.A. Kruse, L.L. Richardson, and V.G. Ambrosia, 1997, "Techniques developed for geologic analysis of hyperspectral data applied to near-shore hyperspectral ocean data,"Proceedings, ERIM 4th International Conference, Remote Sensing for Marine and Coastal Environments, Environmental Research Institute of Michigan (ERIM), Ann Arbor, v. I, p. I-233 - I-246.
H.R. Lang, S.L. Adams, J.E. Conel, B.A. McGuffie, E.D. Paylor, and R.E. Walker, 1987, "Multispectral remote sensing as stratigraphic tool, Wind River Basin and Big Horn Basin areas" Wyoming: American Association of Petroleum Geologists Bulletin, 71, 4, 389-402
R.A. Neville, N. Rowlands, R. Marois, and I. Powell, "SFSI: Canada's First Airborne SWIR Imaging Spectrometer," Canadian Journal of Remote Sensing, Vol. 21, No. 3, pp. 328-336, 1995.
T. Szeredi, K. Staenz, and R.A. Neville, "Automatic Endmember Selection: Part I Theory," Submitted for publication in Remote Sensing of Environment, 1999.
About the Authors
Mr. M. Rajesh Kumar is Research Assistant in the Department of Earth Sciences, Indian Institute of Technology, Bombay. His areas of interest are GIS and remote sensing applications in landslide studies and applied research in natural resource management, with emphasis on water resource management. He can be reached at [email protected].
Dr. T.N. Singh is Associate Professor in the Department of Earth Sciences, Indian Institute of Technology, Bombay. He is an expert in engineering geology and rock mechanics. He can be reached at [email protected].
|
|
©Copyright 2005-2021 by GITC America, Inc. Articles may not be reproduced, in whole or in part, without prior authorization from GITC America, Inc. The original version of this article can be found at: https://www.eomonline.com/EOM_Jun05/article.php?Article=feature04 |