Application of multispectral LiDAR to automated virtual outcrop geology

https://doi.org/10.1016/j.isprsjprs.2013.12.004Get rights and content

Abstract

Terrestrial laser scanning (TLS) is a valuable tool for creating virtual 3D models of geological outcrops to enable enhanced modeling and analysis of geologic strata. Application of TLS data is typically limited to the geometric point cloud that is used to create the 3D structure of the outcrop model. Digital photography can then be draped onto the 3D model, allowing visual identification and manual spatial delineation of different rock layers. Automation of the rock type identification and delineation is desirable, and recent work has investigated the use of terrestrial hyperspectral photography for this purpose. However, passive photography, whether visible or hyperspectral, presents several complexities, including accurate spatial registration with the TLS point cloud data, reliance on sunlight for illumination, and radiometric calibration to properly extract spectral signatures of the different rock types. As an active remote sensing method, a radiometrically calibrated TLS system offers the potential to directly provide spectral information for each recorded 3D point, independent of solar illumination. Therefore, the practical application of three radiometrically calibrated TLS systems with differing laser wavelengths, thereby achieving a multispectral dataset in conjunction with 3D point cloud data, is investigated using commercially available hardware and software. The radiometric calibration of the TLS intensity values is investigated and the classification performance of the multispectral TLS intensity and calibrated reflectance datasets evaluated and compared to classification performed with passive visible wavelength imagery. Results indicate that rock types can be successfully identified with radiometrically calibrated multispectral TLS data, with enhanced classification performance when fused with passive visible imagery.

Introduction

In the past decade, airborne and terrestrial LiDAR (Light Detection and Ranging), also referred to as laser scanning, has gained widespread acceptance and use in measuring and documenting 3D topographic conditions. Laser scanning is used extensively in the private sector for site as-built conditions ranging from complex piping networks in industrial environments to basic civil engineering tasks such as measurements for volume computations. Academic and research organizations have also adopted laser scanning for a wide variety of applications. Examples include seismic event change detection (Oskin et al., 2012), archeological and paleontological study and preservation (Bates et al., 2008, Chase et al., 2011), and forest biomass computations (Lefsky et al., 2002, Lim et al., 2003). The use of high resolution terrestrial laser scanners (TLS) is also well suited for the field of virtual outcrop geology, in which high resolution 3D models of exposed outcrops and cliff sections are analyzed to infer subsurface geology. These investigations are especially prevalent in hydrocarbon reservoir studies where exposed rock layers are used to create 3D models of subsurface rock layers which serve as analogues for understanding fluid flow and storage in similarly deposited reservoirs around the world (Buckley et al., 2010, Pringle et al., 2006).

While laser scanning is very adept at capturing precise 3D models of surface geometry, the highly accurate point clouds generated by LiDAR systems contain little or no information regarding the mineral or chemical composition of the reflecting surfaces. A majority of LiDAR systems collect the strength of the return energy (often referred to as intensity), but this measure is normally un-calibrated and provided at a scale and resolution which varies between instruments. Thus, it is an effective tool for characterizing relative reflectance between materials, but does not directly provide a measure of absolute reflectance without calibration (Franceschi et al., 2009, Hartzell et al., 2013, Kaasalainen et al., 2011, Pfennigbauer and Ullrich, 2010). Even when radiometrically calibrated to determine absolute reflectance, the LiDAR source is nominally from a single wavelength, which may be of limited value when attempting to automatically determine material reflection properties.

To augment the value of laser scanning products, passive digital imagery is sometimes fused with the 3D point cloud data, such as in geological applications where very dense 3D point clouds have been draped with high resolution photography to create detailed virtual rock outcrop representations for enhanced interpretation and analysis (Buckley et al., 2010, Enge et al., 2007, Kurtzman et al., 2009, McCaffrey et al., 2005, Wilson et al., 2009). The color information (RGB) can then be utilized in combination with the laser intensity to extract object properties as in Lichti (2005). More recent work has investigated combining target reflectance properties from high dimensionality passive multispectral or hyperspectral imagery with 3D scan data for enhanced target material identification (Kurz et al., 2011). While this data fusion is promising because it combines the accurate 3D models from the point cloud with the high spectral content of the hyperspectral imagery, there are still a few remaining issues with the optimal combination of these datasets. First, the viewpoint and field of view of the hyperspectral camera and laser scanner are not coincident, and therefore a mapping or draping of the hyperspectral imagery onto the laser point cloud is required, which, if not carefully calibrated, can produce misalignment between the spatial and spectral datasets. Additionally, the hyperspectral datasets are passive, and therefore rely upon the sun for illumination. As a result, calibration for the amount of radiant sunlight incident on the object of interest is required. This calibration for downwelling irradiance is normally undertaken by either placing objects of known reflectance within the image scene, or taking in situ measurements of reflectance of objects within the scene in order to calibrate the hyperspectral imagery (Bachmann et al., 2010).

The requirements of passive hyperspectral imagery for illumination by sunlight and in situ calibration can cause significant problems for its use for virtual outcrop geology, where it is desirable to determine both the structure and material properties of an exposed rock outcrop (Buckley et al., 2010). Many rock outcrops of interest are inaccessible for in situ calibrations, and the vertical nature of a majority of rock outcrops do not provide ideal conditions for uniform illumination by sunlight across the entire outcrop, especially in the case of overhanging walls, or those with a northern exposure (in the northern hemisphere). With these considerations in mind, an ideal tool for virtual outcrop geology would be a hyperspectral laser scanner that afforded the ability to use active illumination to simultaneously spectrally and spatially sample a rock outcrop at high resolution. Recent advances in pulsed lasers, nonlinear fiber optics (Dudley et al., 2006) and avalanche photo diodes (APDs) have made the combination of active illumination and ranging and hyperspectral channel reflectance a possibility with the creation of broadband directional light sources and small, high resolution photo detectors. However, currently available broadband light sources do not have the output power available to allow ranging at longer distances. A number of organizations have demonstrated hyperspectral lasers (Alexander et al., 2013, Hakala et al., 2012, Suomalainen et al., 2011) and multispectral systems (Powers and Davis, 2012, Tan et al., 2005, Wallace et al., 2005, Woodhouse et al., 2011), although most are confined to a laboratory setting. Multiple laser wavelengths have also been used for distinguishing targets by their spectral properties in differential absorption LiDAR for atmospheric studies for several decades (Browell et al., 1998) as well as in bathymetric LiDAR where a near infrared laser is used to detect the water surface and a green laser to detect the benthic layer (Wang and Philpot, 2007).

While the concept of using radiometrically calibrated multispectral LiDAR observations for target identification is not new, only a few research organizations are using the method, often with one-of-a-kind experimental systems. Therefore, it would be advantageous to investigate whether multispectral LiDAR as a tool for virtual outcrop geology can be accomplished using existing commercial hardware. The objective of this study is to demonstrate a practical method of using terrestrial multispectral LiDAR observations, acquired and processed with readily available commercial hardware and software packages, for classification of lithological units in a rock outcrop. The performance of an empirical radiometric calibration is investigated, and classification accuracies achieved from raw and radiometrically calibrated TLS intensities are compared to those obtained from passive visible wavelength imagery, which is often collected in conjunction with TLS data.

Section snippets

Background

In addition to the 3D points computed from the combined ranges and angles measured by LiDAR systems, dimensionless intensity values related to the amount of energy reflected by a target from the incident laser energy are also reported. For a pulsed LiDAR system, assuming a collimated beam and diffusely reflecting surface, the expected amount of reflected photoelectron energy detected can be expressed using a form of the LiDAR link equation (Cossio et al., 2010):ηs=ηqηrEthν·ρλcosαArπR2·[exp(-βe,λ

Instrumentation

Three TLS systems with unique laser wavelengths were selected for use in the study: a Riegl VZ-400, a Leica HDS3000 and a Zoller + Fröhlich (Z + F) IMAGER 5003. Table 1 contains an overview of the specifications for each laser system. For this study, the most important characteristic is that each TLS utilizes a laser at a different wavelength: 0.532 μm (visible green) for the Leica, 0.785 μm (red) for the Z + F, and 1.550 μm (near infrared) for the Riegl.

Radiometric calibration

If reported TLS intensity values scale linearly

Multispectral LiDAR collection

A well-exposed rock outcrop consisting of several distinct sedimentary rocks near Lion Mountain in Kingsland, TX was selected for the study. These rocks are Cambrian, approximately 540 million years old, and were deposited in a complex depositional environment, mainly along a shoreline (Chafetz, 1978). The rock sequence shows an interesting geological history of rising and falling sea level. The outcrop consists of three mappable units, which are numbered from the bottom to the top in Fig. 2.

LiDAR post processing for analysis

Riegl’s RiSCAN PRO software package was used for all filtering and registration tasks for the TLS datasets. Beyond typical filtering tasks performed on all datasets, such as eliminating errant mid-air returns and limiting the spatial extent of the datasets, the Riegl dataset was filtered to include only single-echo returns. This is necessary since the radiometric calibration assumes return intensity values are from single targets. This filter was not necessary for the Leica and Z + F TLS systems,

Results and discussion

Two classification methods were attempted within the ENVI image processing software package. The first method directly classifies the calibrated TLS reflectance images with a MD classifier using the spectral curves measured from the rock samples as endmembers. Given the consistent departure of the calibrated TLS reflectance values from the spectroradiometer reflectance values in the proof of concept test (Section 3.2), the direct classification is not expected to perform as robustly as

Conclusions

A method for target classification based on multispectral LiDAR observations using commercially available hardware and software for potential application to automated virtual outcrop geology was demonstrated. An empirical radiometric calibration of TLS intensity values that accommodated range and intensity non-linearity was found to produce consistent reflectance values, but only closely matched spectroradiometer reflectance values in a limited number of cases. The TLS intensity calibration

Acknowledgements

The authors thank Darren Hauser for his assistance in the field acquisition of the rock outcrop data in Kingsland, TX. The authors also thank the two anonymous reviewers for appraising an earlier version of the text. Their detailed and constructive criticisms improved the overall manuscript. Support for the first author is provided through an appointment to the Student Research Participation Program at the U.S. Army Cold Regions Research and Engineering Laboratory (CRREL) administered by the

References (37)

  • V.V. Alexander et al.

    Power scalable >25 W supercontinuum laser from 2 to 2.5 μm with near-diffraction-limited beam and low output variability

    Opt. Lett.

    (2013)
  • C.M. Bachmann et al.

    Retrieval of substrate bearing strength from hyperspectral imagery during the virginia coast reserve (VCR’07) multi-sensor campaign

    Mar. Geodesy

    (2010)
  • K.T. Bates et al.

    Three-dimensional modelling and analysis of dinosaur trackways

    Palaeontology

    (2008)
  • E.V. Browell et al.

    Differential absorption lidar (DIAL) measurements from air and space

    Appl. Phys. B: Lasers Opt.

    (1998)
  • S.J. Buckley et al.

    Terrestrial laser scanning for use in virtual outcrop geology

    Photogram. Rec.

    (2010)
  • H. Chafetz

    A trough cross-stratified glaucarenite: a Cambrian tidal inlet accumulation

    Sedimentology

    (1978)
  • A.F. Chase et al.

    Airborne LiDAR, archaeology, and the ancient Maya landscape at Caracol, Belize

    J. Archaeol. Sci.

    (2011)
  • T.K. Cossio et al.

    Predicting small target detection performance of low-SNR airborne lidar

    IEEE J. Sel. Top. Appl. Earth Observ. Remote Sens.

    (2010)
  • J.M. Dudley et al.

    Supercontinuum generation in photonic crystal fiber

    Rev. Mod. Phys.

    (2006)
  • H.D. Enge et al.

    From outcrop to reservoir simulation model: workflow and procedures

    Geosphere

    (2007)
  • M. Franceschi et al.

    Discrimination between marls and limestones using intensity data from terrestrial laser scanner

    ISPRS J. Photogramm. Remote Sens.

    (2009)
  • T. Hakala et al.

    Full waveform hyperspectral LiDAR for terrestrial laser scanning

    Opt. Express

    (2012)
  • Hartzell, P.J., Glennie, C.L., Finnegan, D.C., 2013. Calibration of a Terrestrial Full Waveform Laser Scanner, Proc....
  • Jiang, J., Liu, D., Gu, J., Susstrunk, S., 2013. What is the space of spectral sensitivity functions for digital color...
  • S. Kaasalainen et al.

    Study of surface brightness from backscattered laser intensity: calibration of laser data

    IEEE Geosci. Remote Sens. Lett.

    (2005)
  • S. Kaasalainen et al.

    Radiometric calibration of LIDAR intensity with commercially available reference targets

    IEEE Trans. Geosci. Remote Sens.

    (2009)
  • S. Kaasalainen et al.

    Analysis of incidence angle and distance effects on terrestrial laser scanner intensity: search for correction methods

    Remote Sens.

    (2011)
  • D. Kurtzman et al.

    Improving fractured carbonate-reservoir characterization with remote sensing of beds, fractures, and vugs

    Geosphere

    (2009)
  • Cited by (62)

    View all citing articles on Scopus
    View full text