| Remote sensing (RS) is the science of acquiring, processing, and interpreting images and related data (acquired from aircrafts and satellites) that record the interaction between matter and electromagnetic energy. Broad-band multispectral RS has some limitations, such as less band amount, wide band width and rough spectral information expression. Thus a major problem in mineral exploration using multispectral sensors is the insufficient spectral resolution, which doesn't exhibit subtle differences in spectral signatures. The advent of new hyperspectral sensor technology, in terms of both sensor and technique development, has provided opportunity to revisit previous RS approaches to mineral exploration as well as development of improved methods.Hyperspectral RS can provide spectral information of materials usually over several hundreds of narrow contiguous spectral bands, with high spectral resolution on the order of 10 nm or narrower in the visible and infrared wavelengths. As such, hyperspectral data allow identification of specific materials, whereas multispectral data only allow discrimination between classes of materials. As a result, hyperspectral data have obvious advantages in ground object classification and identification that is done by quantitative comparison of known reference spectra (laboratory or field spectra) to unknown image spectra. Thus hyperspectral RS has been one of the most important earth-observing technologies since the 1980s and is a current advanced technology of RS. Imaging spectrometers (hyperspectral sensors) were developed to acquire image data in many narrow spectral bands so that diagnostic absorption features of minerals which are typically 20-40 nm in width could be identified in image spectra to enable mapping. At present several operational airborne imaging spectrometer systems (AVIRIS, HYMAP, CASI, MAIS, etc.) and the first spaceborne hyperspectral sensor (Hyperion) successfully allow surface mineralogical mapping. Despite the fact that the geology community was the first to recognize the importance of hyperspectral technology, nowadays hyperspectral technology has been applied to many fields, including rock and mineral identification, land cover change monitoring, and vegetation type identification, etc.The study area is located in the Pulang porphyry copper deposit and its periphery area, Zhongdian County of Yunan province, China. Hyperspectral image data of this area, from the American satellite EO-1's Hyperion, was acquired in November, 2003. The field spectra were collected with FieldSpectral Pro, an Analytical Spectral Devices (ASD) spectrometer. Some mineral spectra used come from the current famous spectra libraries, such as USGS,JPL and JHU. To accomplish prospecting-oriented approaches to information extraction of rocks and minerals using hyperspectral RS data, this paper analyses the regional geological background, and district geological features of the study area, and then studies some steady absorption feature characteristics of field rocks and minerals spectra. The main research results are as follows:1. The preprocessing of EO-1 Hyperion hyperspectral data includes: converting Hyperion data to absolute radiance, restoring bad lines and removing vertical stripes, correcting for atmospheric scattering and correcting geometric distortions. The hyperspectral data of the study area have been successfully preprocessed and the methods of bad lines' restoral and vertical stripes' removal are implemented with the aid of a VC++ programe.2. Alteration minerals in the study area mainly include limonite, pyrite, illite, kaolinite, muscovite, montmorillonite, chlorite, tremolite, calcite and dolomite. To extract alteration information from the hyperspectra image, the spectral features of these minerals are analyzed, and their laboratory spectra and the method of SAM (Spectral Angle Mapping) are used. The extracted alteration information is obviously concentrated in Are-Pulang-Langdu area which is at altitude of above 4200 meters and has little vegetation coverage. But in other areas, especially some vegetation covering areas, alteration information can't be availably extracted. The results show that these used laboratory minerals spectra aren't very appropriate to spectra matching with image spectra because minerals spectra aren't steady under the influence of field environment and noisy surface properties of ground objects. At the same time, the uncertainty of the used thresholds in information extraction often results in different results, thus a map is indispensably used to demonstrate match degree between each pixel spectrum and its reference spectrum so that the results can be easily accepted or rejected.3. Spectral variation characteristics of rocks and minerals under the influence of different field environments are analyzed, and the steady spectral characteristics (spectral overall shape and absorption-band position) are obtained. The two methods based on spectral vector: SAM and SCM (Spectral Correlation Mapping) are applied to field ground object spectra of the study area. The results indicated that most objects have a great matching value with their similar objects, and SAM has a smaller identification degree than SCM. Spectral similarity measures based on absorption features are studied, and the result shows that the identification of field rocks and minerals based on spectral absorption features can be done by quantitative absorption-band positions' comparison of known reference spectra (laboratory or field spectra) to unknown image spectra.4. This paper presents a new approach to object identification of hyperspectral image based on the spectral exhaustive method (EXM), and the approach has been implemented. The key techniques and methods in this approach include: spectral library and its management, the normalizing processing of continuum removal, and Mapping and analyzing functions of identification results. Mapping of identification results is accomplished using a color (RGB) map and a gray-grade map. The color map is generated using the reference spectra with different colors so that all pixels can be shown as the identified type. The gray-grade map is created based on the correlation coefficient between each pixel spectrum and its reference spectrum, which demonstrates the degree of match. These two maps are evaluated in order to select suitable reference spectra for the study and judge the accuracy of the object identification results.The approach was applied to an EO-1 Hyperion hyperspectral image of the study area. The results indicated that the fourteen reference spectra are suitable for this study and the identification results are satisfactory. In eighteen field ground objects examined, although some of their corresponding pixels have a small correlation coefficient, the identified objects corresponding to the pixels are basically identical to the field ground objects. The identification results of A and B area (outside of experiment area) show that as long as the reference spectra are associated with object spectra in the area, each pixel spectrum of the area can find out its most similar reference spectrum. But some correlation coefficients in the gray map are small because of the influence of field environment (vegetation, cloud, snow etc.). On the other hand, it is known that field objects (such as vegetation, soil) have a great influence on rock spectra. Thus it is necessary to think about environment factors which influence identification of alteration rocks in prospecting.The continuum-removed spectra can enhance their absorption features, and it is especially obvious to most vegetation spectra except for the object spectra with flat overall shape. The identification results of the continuum-removed image show that the normalizing processing of continuum removal isn't advantageous to the rocks' identification in the hyperspectral image with intensive environment information.5. In order to more accurately identify ground objects with similar spectral overall shape, it is important to analyze the absorption-band positions. This paper presents a development in EXM which is mainly based on analysis of the absorption-band positions, and get a flow for object identification based on the spectral overall shape character and local absorption feature. The key techniques and methods in the identification include: determination of absorption-band offset rang, and spectral matching algorithm of absorption-band position. Determination of absorption-band offset range is done via three steps:â‘ For each known ground object, many spectra are obtained in the field. The cross correlogram spectral matching (CCSM) is used to acquire the maximal overall offset from these spectra.â‘¡The algorithm of absorption-band positions automatically obtained is implemented based on the method of continuum removal. The absorption-band positions (corresponding to the same material) in these spectra are analyzed to acquire the maximal position offset.â‘¢An offset range is got on the basis of the maximal overall offset and the maximal position offset. This offset range is taken as a constraint on the matching process. Spectral matching of absorption-band position is accomplished by satisfying the two criterias:â‘ Wavelength difference between the corresponding absorption-band positions in the two spectra is within the offset range.â‘¡The correlation coefficient between the two spectra, calculated within each absorption-band width of the reference spectrum, is greater than a specified value.The offset results of five typical field object spectra show that overall offsets of object spectra are small in the study area and the offset of absorption-band position is mainly caused by the offset of the position. The application results show that the three ground objects (their serial numbers are R11, R12 and R13) have a small a small correlation coefficient in the overall shape between them and their reference spectra, but their similarity in absorption features (including positions of absorption bands) can provide valuable information for the identification of objects. Thus the method of object identification based on the spectral overall shape character and local absorption feature can get more accurate identification results than the method based on spectral overall shape alone.6.The identification results of Pulang porphyry copper deposit area using this area's field spectra are thought to be a more accurate identification results, for which there are two resasons: on one hand, the method based on the spectral overall shape character and local absorption feature takes the object's physical and chemical properties, and spectral overall shape into consideration. On the other hand, the field spectra own the characteristics of Pulang porphyry copper deposit area, thus they are suitable for the object identification of this area. The hyperspectral image of Pulang porphyry copper deposit area and its periphery area is identified using the fourteen spectra, and then percents of the two areas' pixel numbers within different correlation coefficients are analyzed. The results show that the fourteen field spectra can be used in the identification of periphery area and the examination of the actual objects indicated that the identification is effective.The hyperspectral image of Pulang porphyry copper deposit area and its periphery area is identified by dividually using fourteen field spectra (deposit area) and eighteen field spectra (deposit area and its periphery area), then the identification results indicated that field alteration rocks in the Pulang porphyry copper deposit area can be used in the identification of periphery area, which can provide a clue to alteration information extracted in unknown area using the field spectra of known typical deposit area. At the same time, alteration information map of the study area is obtained. The results shown in the map indicated that there is much obvious alteration information in the Disua-Langdu-Zhuoma-Bidu area and Disua-Songnuo-Pulang area which are in accord with the areas of porphyry bodies in geological map of the study area, and there is obvious alteration information in Ousaila area and Are area which are almost in accord with the areas of porphyry bodies. All the alteration information can be found in the field examination.The main innovations of this study are as follows:1. The steady spectral characteristics (spectral overall shape and absorption-band position) are obtained by analyzing spectral variation characteristics of field rocks and minerals.2. This paper presents a new approach to object identification of hyperspectral image based on the spectral exhaustive method, and the approach helps us select suitable reference spectra for the study and judge the accuracy of the object identification results.3. This paper presents a development in EXM which is based on analysis of mainly the absorption-band positions, and get a flow for object identification based on the spectral overall shape character and local absorption feature.4. It is proved that the field alteration rocks in the Pulang porphyry copper deposit area can be used in the identification of periphery area, which can provide a clue to alteration information extracted in unknown area using the field spectra of known typical deposit area.Nowadays some methods of information extraction are being studied using hyperspectral image and object spectra in the application of hyperspectral RS to the mineral exploration. Spectral modeling, spectral matching and geological mapping become important and hot techniques in the application. Absorption-band parameters include the position, depth, width, and asymmetry of the feature, etc. This paper obtains steady spectral characteristics of field spectra in the study: spectral overall shape and absorption-band position, and then studies the prospecting-oriented approaches to information extraction of rocks and minerals. This research accomplishs the prospecting-oriented approaches to information extraction of rocks and minerals using hyperspctral RS data, these approaches are effective in this study area and will be tested in other areas for a wider application.
|
PDF Full Text Request