A comparative approach on TIR and VNIR-SWIR datasets of ASTER instrument for lithological mapping in Neyriz ophiolite zone, SW Iran4.2. Spectral angle mapping (SAM) The Spectral Angle mapping (SAM) is a physically-based spectral classification that uses an n-dimensional angle to match pixels to reference spectra (Kruse et al., 1993). The algorithm determines the spectral similarity between two spectra by calculating the angle between the spectra, treating them as vectors in a space with dimensionality equal to the number of bands. This technique, when used on calibrated reflectance data, is relatively insensitive to illumination and albedo effects. Poorly illuminated pixels would fall closer to the origin than pixels with the same spectral signature but greater illumination. The SAM algorithm generalizes this geometric interpretation to hyper-dimensional space and permits rapid mapping of the spectral similarity of image spectra to reference spectra (Kruse et al., 1993), using the arccosine of the dot products of image and reference spectra as vectors in x-dimensional space. For each reference spectrum chosen in the analysis of an image, the spectral angle is determined for every image spectrum. The outputs of the algorithm are a classification image displaying the best SAM match and a RULE image for each end-member showing the spectral angles in radians between each spectrum in the image and the reference spectrum. Pixels with smaller spectral angles are darker in rule images. Density slice thresholds are used to empirically determine those areas that most closely match the reference spectrum while retaining spatial coherence.  Figure 5: Output image of SAM analysis on the spectrum of pillow lavas. The red areas show the best coincidence with the pillow lavas outcrops.  Figure 6: Output image of SAM analysis on the spectrum of gabbro. The red areas show the best coincidence with the gabbro outcrops.  Figure 7: Output image of SAM analysis on the spectrum of dunite. The red areas show the best coincidence with the peridotite outcrops.  Figure 8: Output image of SAM analysis on the spectrum of marble. The red areas show the best coincidence with the marble outcrops. The spectra of collected rock samples and the spectra of selected rocks from USGS spectral library were used as the end members for lithological enhancements. The algorithm was implemented on IARR calibrated SWIR dataset and the output RULE images were validated using the geological map and field evidences. The spectra of pillow lavas and sheeted dykes, gabbro and dunite were selected from USGS spectral library, while the spectra of marble was selected from the collected samples. Examination of density slice ranges of Rule image of pillow lavas showed that areas with thresholds 0.003-0.01 and 0.01-0.015 radians coincided the outcrops of pillow lavas at the northern part of main ultramafic mass (Figure 5). The output image for gabbro (Figure 6) shows the relevant areas with thresholds 0.0024-0.01 and 0.01-0.02 radians at the northern and eastern parts of the study area, coincided with the same exposures in geological map. Examination of density slice ranges of Rule image of dunite showed that areas with threshold 0.0047-0.025 and 0.025-0.035 radians were coincided to the main peridotite exposures of the Neyriz ophiolite zone (Figure 7). The output Rule image of marble spectrum shows the ranges of 0.0024-0.01 and 0.01-0.02 radians for the same exposures, mainly in Tang-e-hana and Ghalae-Bahman marble mines (Figure 8). 4.3. Minimum Noise Fraction Minimum Noise Fraction (MNF) analysis identifies the locations of spectral signature anomalies. This process is of interest to explorationists because spectral anomalies are often indicative of alterations. The minimum noise fraction (MNF) transform is used to determine the inherent dimensionality of image data to segregate noise in the data and to reduce the computational requirements for subsequent processing (Boardman and Kruse, 1994). This method is similar to principal component (PC) analyses that have been used for a long time in multispectral image processing, but involves an extra preceding step. The MNF procedure was examined on the TIR dataset of ASTER. The first output image enhances marbles and Orbitolina limestones in bright pixels, peridotites in grey and other rock units in dark pixels. The second component enhances radiolarites in bright, marbles in grey and basic alluvium in dark pixels. The third component also enhances radiolarites in bright pixels, but 4th and 5th components are mostly noisy images and do not enhance any rock unit. Gabbros, pillow lavas and diabasic dykes are not enhanced in any component. RGB color composite of three first components discriminates main rock units in different colors, but the gabbroic rocks are not discriminated from peridotites (Figure 9). This procedure was also implemented on VNIR-SWIR bands. Rock units discriminated in a color composite image generated from first three components were similar to those enhanced in previous colored image.  Figure 9: A color composite image produced from MNF1, MNF2, and MNF3 components, output results of performing MNF algorithm on TIR dataset of ASTER. 4.4. Matched Filtering (MF) Matched filtering is a technique which performs partial unmixing, finding the abundances of user defined endmembers. Not all of the endmembers in the image need to be known. This technique maximizes the response of the known endmember and suppresses the response of the composite unknown background, thus “matching” the known signatures. It provides a rapid means detecting specific materials based on matches to library or image endmember spectra and does not require knowledge of all endmembers within an image scene. Bands 4-9 have been processed by MF in order to map marbles and Orbitolina limestone, based on the marble spectrum examined from study area samples, and the result was compared to the geological map. The output image enhances all the outcrops of marbles in bright pixels. The same procedure was tested on TIR dataset of ASTER for dunites, gabbros, and pillow lavas using the spectra of USGS. Dunites and gabbros were enhanced in bright pixels, while pillow lavas were discriminated in dark pixels. SUMMARY AND CONCLUSIONS The recent development of truly multi-spectral remote sensing systems such as the ASTER instrument potentially offers geologists a cost-effective solution to expensive and time-consuming regional mineral exploration and geological mapping. Subtle spectral reflectance differences recorded in both the VNIR and SWIR wavelength regions are an important basis for identifying specific individual minerals and mineral groups and is a significant advancement over earlier sensors. Digital image processing techniques such as PCA, spectral angle mapping, and matched filtering were implemented on both VNIR-SWIR and TIR datasets of ASTER using the spectral characteristics of rock units exposed at the Neyriz ophiolite zone. The PCA technique is very effective for discriminating rock units where spectral information is not available and/or ground truth is limited. Performing PCA approach on SWIR dataset enhances marbles, basic alluviums and gabbros. A RGB color composite image generated from PCA outputs of thermal bands 10, 12, and 13 discriminates most various rock types in different colors. Examining the output Rule images, derived from performing spectral angle mapping on dunite, gabbros, pillow lavas and marble spectra showed that the enhanced areas were coincided on the relevant rock exposures in geological map and as observed in field. Although main rock types are showed in a RGB color composite generated from components 1, 2, and 3, of minimum noise fraction algorithm, gabbros are not discriminated from peridotites in the same colored image. Performing matched filtering on SWIR and TIR datasets using the specra of marbles, dunites, pillow lavas and gabbros results in enhancement of these rock types. The results obtained from this work on the ASTER data of Neyriz ophiolite zone and adjacent outcrops show that the main lithologies can be mapped using the compositional information derived from spectral characteristics of the samples. It is concluded that although using VNIR-SWIR data enhances some of the ophiolitic units, but spectral similarities between rock units prevents the mapping of all the lithologies, however, these datasets show the carbonate units at the northeastern and southern the study area. It is also suggested that the TIR data of ASTER could successfully discriminate mafic-ultramafic rock units as well as radiolarites, the main lithological units of the ophiolite zone. ACKNOWLEDGEMENTS I would like to thank the Research Vice Chancellor of Shiraz University, Iran, for providing the funds and field logistics. I express my gratitude to Dr. Marc Goossens (Director of “Geosense mineral exploration company”, the Netherlands) for his collaboration in accessibility to the ASTER data and examining the spectra of rock samples. I also thank Miss. N. Mazhari for her assistance in data processing. REFERNCES
Page 3 of 3
| Previous | |