Integration of ground-based hyperspectral and lidar scanning in virtual outcrop geology
MetadataShow full item record
The potential to visualize and analyse geological outcrops in a 3D environment has made terrestrial laser scanning (TLS) a standard method in geological field studies. Lidar models can be integrated with high resolution photographs to generate photorealistic 3D models, also referred to as virtual outcrop models (VOMs) in geological applications. However, the extraction of mineralogy and geochemical variations from VOMs is limited to the visible light of the photographs and to the single spectral band provided by the laser sensor. Imaging spectrometry applied from airborne and spaceborne platforms is an established method for the regional mapping of mineralogy and lithology, utilising the interaction of solar radiation with the Earth’s surface. Many minerals and rocks can be mapped and analysed in a non-contact manner by utilizing their diagnostic absorption properties within the visible and particularly within the infrared spectral range. The aim of this research is to apply imaging spectrometry with a ground-based instrument to enable mineralogical and lithological analysis of near-vertical outcrop sections. The terms ground-based and close-range are used to indicate a nearhorizontal setup, as opposed to the nadir view found in airborne and spaceborne applications. A workflow has been developed to integrate hyperspectral classifications with 3D lidar models, to compliment VOMs with reliable information about the mineralogy and geochemical variations in the outcrop. The workflow includes data acquisition, spectral and photogrammetric processing of the hyperspectral images, data integration and classifying VOMs utilising hyperspectral image products. A newly developed hyperspectral imager designed as a compact and lightweight instrument, and therefore practical for field applications, has been used. The HySpex SWIR-320m sensor operates within the short wave infrared light (SWIR) with a spectral range between 1.3-2.5 μm. The spectral data were processed with methods primarily developed for airborne and spaceborne applications. All images showed a significant amount of image artefacts, mainly related to the irregular illumination-viewing geometry and bad pixels. While image nonuniformities such as bad pixels are a common problem in pushbroom scanning, other artefacts such as intensity gradients in along-track direction are exacerbated by the close-range scanning and panoramic image geometry. Applying different nonuniformity corrections, image artefacts were minimized but could not be completely removed. For materials with 50% reflectance a signal to noise ratio better than 70:1 was achieved. Atmospheric corrections were performed utilising an Empirical Line correction, based on two reference spectra measured from calibrated Spectralon panels which were placed in the image scene. Due to a restricted view of the upper hemisphere in close-range scanning, the obtained reflectance values need to be considered as conic-directional reflectance. To separate and remove image noise Maximum Noise Fraction transform was applied. Spectral classification and mapping was performed using different approaches including band ratios, Spectral Angle Mapping, Spectral Feature Fitting and Mixture Tuned Match Filtering. Based on a cylindrical camera model, VOMs could be integrated and textured with hyperspectral image products with an accuracy of one image pixel. Two case studies from different geological settings were carried out, to demonstrate how close-range hyperspectral imaging can help improve the analysis of vertical outcrops. In the Pozalagua quarry (Spain), hydrothermal dolomitized limestone of Cretaceous platform-slope carbonates have been spectrally mapped. Despite very similar chemical and spectral properties, different dolomite and limestone types, as well as calcite could be distinguished and mapped in the outcrop. Spectral differences of two main dolomite types could be related to different manganese and iron contents, as confirmed by chemical analysis. Although detailed spectral analysis was disturbed by surface weathering products, dolomite and limestone were also mapped on weathered surfaces. A limestone unit initially missed by conventional field observations, due to similar visual appearance compared to the surrounding carbonates, was clearly identified and mapped by spectral means. The second field area was Garley canyon (Utah, USA), where a shallow marine, shoreface succession was studied. Carbonate and clay abundances were determined to map and quantify carbonate concretions, and to map siltstone and sandstone in the outcrop. Carbonate concretions have implications for reducing porosity and permeability in shallow marine sandstones. Results show that close-range imaging spectrometry can provide reliable qualitative and quantitative information about the mineral-chemical composition of exposed surfaces. Further research is required to improve the nonuniformity, atmospheric and topographic correction of the spectral images and to adjust the processing to the close-range scanning and image geometry. However, the method can be adapted to other applications in which the collection and analysis of chemical surface composition and geometric information is required, such as in mining, building damage assessment or in forestry for canopy analysis. With an increased availability of lightweight hyperspectral imagers it is expected that close-range imaging spectrometry will become a sub-discipline in remote sensing, and a standard method in field-based geoscience studies.
Has partsPaper 1: Kurz, Tobias H.; Buckley, Simon J. and J John A. Howell, Close-range hyperspectral imaging and lidar scanning for geological outcrop analysis: Workflow and methods. Draft version. Full text not available in BORA.
Paper 2: Kurz, Tobias H.; Buckley, Simon J.; Howell, John A. and Danilo Schneider, Integration of panoramic hyperspectral imaging with terrestrial lidar data. Draft version. Full text not available in BORA.
Paper 3: Kurz, Tobias H.; Buckley, Simon J., and John A. Howell, Remote lithological mapping of geological outcrops using close range hyperspectral imaging. Draft version. Full text not available in BORA.
Paper 4: Kurz, Tobias H.; Dewit, Julie,; Buckley, Simon J.; Thurmond, John B.; Hunt, David W. and Rudy Swennen, Hyperspectral image analysis of different carbonate lithologies (limestone, karst, hydrothermal dolomites): the Pozalagua Quarry case study (Cantabria, NW Spain). Accepted version. Accepted for publication in Sedimentology. Copyright Wiley-Blackwell. Full text not available in BORA due to publisher restrictions.
PublisherThe University of Bergen
Copyright the author. All rights reserved