Terrestrial lidar and hyperspectral data fusion products for geological outcrop analysis
Highlights
► Lidar and hyperspectral fusion gives geoscientists new ways to map outcrop content. ► Multitextured 3D photorealistic models aid communication and validation of results. ► Classification vectors are mapped to 3D space to aid quantitative analysis. ► Accuracy of registration and classification can be assessed using the data fusion.
Introduction
The combination of data and products derived from multiple spatial and geophysical acquisition sensors is common in the study of geological outcrops. Data sources are often complementary, allowing more information to be extracted than from a single component technique (e.g. Jones et al., 2009). The past decade has seen the rapid adoption of digital spatial measurement techniques in field geology. Global Navigation Satellite Systems, photogrammetry, airborne and terrestrial laser scanning (lidar), digital elevation models (DEMs), and remote sensing imagery have all been used to provide a precise framework for digital mapping and interpretation, at multiple scales (McCaffrey et al., 2005). The integration of these geometric data with traditional geological field measurements (e.g. sedimentary logs, photographs and structural measurements) and geophysical data (e.g. ground-penetrating radar and petrophysical measurements), in advanced modelling and visualisation software, has changed the way that many geological problems are addressed (Jones et al., 2009).
The integration of terrestrial laser scanning with digital imaging has become common in outcrop geology, as well as the wider geoscience discipline. Laser scanning is now a widespread means of obtaining precise and high resolution three-dimensional (3D) topographic information, with high efficiency and ease of use (Bellian et al., 2005, Buckley et al., 2008, Hodgetts, 2009). Most laser scanners suitable for outcrop-scale applications (c. 10 m–1 km) obtain 3D point positions by measuring the range between the instrument and a target surface, based on the time of flight of a laser pulse, together with its direction. The strength of the returned laser pulse, commonly referred to as intensity, is also recorded (Höfle and Pfeifer, 2007). Measurements are made many thousands of times per second, resulting in the formation of a dense point cloud that accurately describes the outcrop surface in great detail, and may be displayed using the recorded intensities. However, the inherently discrete nature of a point cloud makes it difficult to interpret without ancillary imagery (Buckley et al., 2008). In addition, geological features may have a minimal 3D signature, but may be easily apparent in the 2D images as colour or edge information. Therefore, integration of the laser point cloud with digital imagery gives an additional continuous data source that can enhance geological interpretation. This is performed using an integrated camera, or by registering separately captured imagery to the lidar scans. In both cases, point clouds can be assigned red, green and blue (RGB) values by projecting each point into an appropriate image (e.g. White and Jones, 2008), though the discontinuous nature of the lidar point cloud often remains a limitation during visualisation.
Conversion of the lidar point cloud to a triangular mesh makes the outcrop representation continuous. Adjacent points are connected by triangle edges, and the digital photos can then be textured onto the mesh (e.g. El-Hakim et al., 1998, Bellian et al., 2005). The result is a photorealistic model (Xu et al., 2000) that has facilitated interpretation, quantification and education in many reported projects (e.g. Bellian et al., 2005, Fabuel-Perez et al., 2009, Buckley et al., 2010, Enge et al., 2010). However, such models provide little quantitative information on the mineral and chemical composition of the outcrop. Knowledge of the distribution of the minerals and materials in an outcrop is valuable for assessing connectivity of rock bodies, but may be difficult to determine remotely, requiring laborious spot sampling in accessible areas. A number of studies have attempted to use spectral approaches to remotely classify materials at close-range. These have included the analysis of lidar intensity to separate different lithologies (Bellian et al., 2005, Franceschi et al., 2009, Burton et al., 2011) and surface properties (Pesci and Teza, 2008, Nield et al., 2011), the use of multispectral photographs (e.g. Lerma, 2001), the combination of lidar intensity and RGB imagery (Lichti, 2005), and the use of multispectral lidar sensors (Hemmleb et al., 2006). Despite the reported successes, current laser scanners are limited by their spectral sensitivity, typically a single narrow wavelength in the visible or near-infrared part of the electromagnetic spectrum (Höfle and Pfeifer, 2007). This precludes the possibility for surface classification where complex material compositions exist.
An enhanced spectral range and resolution is offered by close-range hyperspectral imaging. Hyperspectral sensors measure many narrow spectral bands across an extended part of the electromagnetic spectrum, allowing a near-continuous reflectance curve to be derived per image pixel (see e.g. van der Meer and de Jong, 2001). High spectral resolution permits subtle variations in surface composition to be quantitatively analysed, even at sub-pixel levels (Keshava and Mustard, 2002). In geology, hyperspectral imaging has been successfully employed using airborne and spaceborne sensors for decades, for regional mapping and mineral prospecting, detection of hydrocarbon seeps, and interplanetary exploration (e.g. van der Meer and de Jong, 2001, Bellian et al., 2007, Bowen et al., 2007, Griffes et al., 2007). Lightweight hyperspectral sensors are now available, making the method applicable from the ground in field geology (Kurz et al., 2008, Kurz et al., 2012, Murphy et al., 2012). These sensors negate the past problems associated with imaging outcrop surfaces – typically having near-vertical orientation – from nadir platforms where field of view and spatial resolution are not optimal.
The aim of the current paper is to highlight the potential of combining terrestrial laser scanning and close-range hyperspectral imaging in outcrop geology. The focus is on the products that can be obtained from the fused data, along with means of assessing the quality of both the geometric fit and the hyperspectral classifications. This is achieved using an empirical workflow developed using a specific hyperspectral device, where the imaging geometry dictates the method chosen for integration with the lidar data. The instrumentation and processing steps are briefly outlined for both component techniques, as well as the procedure for co-registering the hyperspectral image with the lidar data. The paper presents a visualisation method that allows multiple hyperspectral processing products to be displayed with the photorealistic lidar model, and outlines quantitative analysis that can be performed following integration. Benefits of the synergy are presented throughout the paper using example outcrop datasets.
Section snippets
Instruments
The proposed integration method is developed using data from a Riegl LMS-Z420i terrestrial laser scanner (Riegl, 2011) and a HySpex SWIR-320 m hyperspectral imager (NEO, 2011). The former is one of several time-of-flight-based laser scanners currently available that share similar specifications, and that are suitable for geological outcrop studies. In this project the scanner is equipped with a calibrated Nikon D200 digital camera, mounted rigidly on top of the instrument body. Captured digital
Hyperspectral registration
Hyperspectral images, like conventional photographs, are essentially dimensionless, needing the correct exterior orientation to be found to relate pixel coordinates to a 3D coordinate system. This information may be recovered using control points that can be measured within an image. The photogrammetric collinearity equations describe the straight line formed between a 3D point on the object and the conjugate point within the captured image, passing through the camera centre. In space
Data fusion products
Once the hyperspectral imagery has been successfully registered to the lidar data, it can be exploited in a spatially meaningful way, allowing geoscientific analysis to be carried out. This section presents some of the possibilities for using the two datasets in combination, for visualising and communicating the results of hyperspectral classification, for extracting information on material positions and areas, and for validating the input data, registration accuracy and success of image
Conclusion
While terrestrial laser scanning has been used extensively for analysing outcrop topography, the integration with ground-based hyperspectral imaging gives geoscientists new ways to analyse outcrop composition. Hyperspectral imaging allows the distribution of mineralogy and lithology to be mapped, even where subtle chemical variations exist. To use hyperspectral images in a spatially meaningful way, data fusion between 2D images and a 3D reference surface must be performed with high accuracy, as
Acknowledgements
Parts of this work were funded by the Research Council of Norway's Petromaks programme (projects 163264 and 176132) and sponsoring companies. Statoil ASA is thanked for supporting fieldwork in Spain. The authors thank Norsk Elektro Optikk AS and Riegl GmbH for continued hardware and software support. Richard Jones and one anonymous reviewer are thanked for their constructive comments.
References (43)
- et al.
A multi-sensor approach to creating accurate virtual environments
ISPRS Journal of Photogrammetry and Remote Sensing
(1998) - et al.
Discrimination between marls and limestones using intensity data from terrestrial laser scanner
ISPRS Journal of Photogrammetry and Remote Sensing
(2009) - et al.
Correction of laser scanning intensity data: data and model-driven approaches
ISPRS Journal of Photogrammetry and Remote Sensing
(2007) - et al.
Integration of regional to outcrop digital data: 3D visualisation of multi-scale geological models
Computers & Geosciences
(2009) Delaunay refinement algorithms for triangular mesh generation
Computational Geometry: Theory and Applications
(2002)- et al.
Digital outcrop models: applications of terrestrial scanning lidar technology in stratigraphic modeling
Journal of Sedimentary Research
(2005) - et al.
Analysis of hyperspectral and lidar data: remote optical mineralogy and fracture identification
Geosphere
(2007) - et al.
Reflectance spectroscopic mapping of diagenetic heterogeneities and fluid-flow pathways in the Jurassic Navajo Sandstone
AAPG Bulletin
(2007) - et al.
Terrestrial laser scanning in geology: data acquisition, processing and accuracy considerations
Journal of the Geological Society
(2008) - et al.
Terrestrial laser scanning for use in virtual outcrop geology
Photogrammetric Record
(2010)
Lidar intensity as a remote sensor of rock properties
Journal of Sedimentary Research
Feature-driven multilayer visualization for remotely sensed hyperspectral imagery
IEEE Transactions on Geoscience and Remote Sensing
High spectral resolution reflectance spectroscopy of minerals
Journal of Geophysical Research, B: Solid Earth
The geometry and internal architecture of stream mouth bars in the Panther Tongue and the Ferron Sandstone members, Utah, USA
Journal of Sedimentary Research
A new approach for outcrop characterization and geostatistical analysis of a low-sinuosity fluvial-dominated succession using digital outcrop models: upper Triassic Oukaimeden Sandstone Formation, central High Atlas, Morocco
American Association of Petroleum Geologists Bulletin
A transformation for ordering multispectral data in terms of image quality with implications for noise removal
IEEE Transactions on Geoscience and Remote Sensing
Geologic and spectral mapping of etched terrain deposits in northern Meridiani Planum
Journal of Geophysical Research
Multispectral data acquisition and processing techniques for damage detection on building surfaces. International Archives of the Photogrammetry
Remote Sensing and Spatial Information Sciences
LiDAR in the environmental sciences: geological applications
Spectral unmixing
IEEE Signal Processing Magazine
Geological outcrop modelling and interpretation using ground based hyperspectral and laser scanning data fusion
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Cited by (100)
Application of combined terrestrial laser scanning and unmanned aerial vehicle digital photogrammetry method in high rock slope stability analysis: A case study
2022, Measurement: Journal of the International Measurement ConfederationAccurate hyperspectral imaging of mineralised outcrops: An example from lithium-bearing pegmatites at Uis, Namibia
2022, Remote Sensing of EnvironmentCitation Excerpt :Snyder et al. (2016) merged visible and near-infrared (VNIR) ground-based hyperspectral imagery with 3D information to map sedimentary structures. Kurz et al. (2013) and Buckley et al. (2013) developed a workflow for the integration of terrestrial light detection and ranging (LiDAR) point clouds and ground-based hyperspectral scanning. This was one of the first times terrestrial hyperspectral imagery was draped over a LiDAR derived 3D model in order to visualise the hyperspectral data in an accurately scaled representation.
Modified-mean-shift-based noisy label detection for hyperspectral image classification
2021, Computers and GeosciencesCitation Excerpt :Hyperspectral images (HSI) are used in many applications because of the many geological and environmental benefits that provide in identifying the Earth’s surface such as agriculture and monitoring vegetation (Yao et al., 2016; Kang et al., 2018c), military and defense (Kang et al., 2017; Cheng et al., 2017a), and security (Kang et al., 2018b; Cheng et al., 2016), civilians aspects, mineral exploration and etc. Processes are performed on HSIs, including HSI denoising, classification (Petropoulos et al., 2012), image segmentation (Chen et al., 2018), target detection, end-member extraction (Penn, 2002) and data fusion (Buckley et al., 2013). Also, softwares have been developed in this area, including (Biehl and Landgrebe, 2002; Gege, 2014; Zhang et al., 2009).
Multi-scale, multi-sensor data integration for automated 3-D geological mapping
2021, Ore Geology ReviewsCitation Excerpt :However, many rock-types have indistinguishable signatures in the visible part of the spectrum, so visible near infrared (VNIR; 400 to 1000 nm) or short-wave infrared (SWIR; 1000 to 2500 nm) hyperspectral data are needed. A significant body of literature exists on geological mapping with hyperspectral data (see van der Meer et al., 2012, for a review) but, as this field has emerged from the satellite remote-sensing community, high fidelity 3-D information is typically unavailable and rarely included (Fraser et al., 2006; Kurz et al., 2011; Buckley et al., 2013; Lorenz et al., 2018; Salehi et al., 2018; Kirsch et al., 2018, 2019). Recent studies have bridged the gap between the digital outcrop and hyperspectral approaches by fusing high-resolution point clouds with high spectral resolution imagery to create ‘hyperclouds’ (Buckley et al., 2013; Kirsch et al., 2019, 2018; Kurz et al., 2011; Lorenz et al., 2018; Salehi et al., 2018).
Low-cost UAV applications in dynamic tropical volcanic landforms
2021, Journal of Volcanology and Geothermal ResearchMapping Lithium-Bearing Pegmatite with Short-Wave Infrared (SWIR) Hyperspectral Imaging of Jingerquan Li-Be-Nb-Ta Pegmatite Deposit, Eastern Tianshan, NW China
2024, IEEE Transactions on Geoscience and Remote Sensing