ArticlePDF Available

Abstract and Figures

Two centuries ago William Smith produced the first geological map of England and Wales, an achievement that underlined the importance of mapping geological contacts and structures as perhaps the most fundamental skill set in earth science. The advent of large digital datasets from unmanned aerial vehicle (UAV) and satellite platforms now challenges our ability to extract information across multiple scales in a timely manner, often meaning that the full value of the data is not realised. Here we adapt a least-cost-path solver and specially tailored cost-functions to rapidly extract and measure structural features from point cloud and raster datasets. We implement the method in the geographic information system QGIS and the point cloud and mesh processing software CloudCompare. Using these implementations, the method can be applied to a variety of three-dimensional (3D) and two-dimensional (2D) datasets including high-resolution aerial imagery, virtual outcrop models, digital elevation models (DEMs) and geophysical grids. We demonstrate the algorithm with four diverse applications, where we extract: (1) joint and contact patterns in high-resolution orthophotographs; (2) fracture patterns in a dense 3D point cloud; (3) earthquake surface ruptures of the Greendale Fault associated with the Mw7.1 Darfield earthquake (New Zealand) from high-resolution light detection and ranging (LiDAR) data, and; (4) oceanic fracture zones from bathymetric data of the North Atlantic. The approach improves the objectivity and consistency of the interpretation process while retaining expert-guidance, and achieves significant improvements (35–65 %) in digitisation time compared to traditional methods. Furthermore, it opens up new possibilities for data synthesis and can quantify the agreement between datasets and an interpretation.
Content may be subject to copyright.
1
Rapid, semi-automatic fracture and contact mapping for point
clouds, images and geophysical data
Samuel T. Thiele1, Lachlan Grose1, Anindita Samsu1, Steven Micklethwaite1, Stefan A.
Vollgger1, Alexander R. Cruden1
1School of Earth, Atmosphere and Environment, Monash University, Melbourne, 3800, Australia
5
Correspondence to: Samuel T. Thiele (sam.thiele@monash.edu)
Abstract. Two centuries ago William Smith produced the first geological map of England and Wales, an
achievement that underlined the importance of mapping geological contacts and structures as perhaps the most
fundamental skill set in earth science. The advent of large digital datasets from unmanned aerial vehicle (UAV)
and satellite platforms now challenges our ability to extract information across multiple scales in a timely manner,
10
often meaning that the full value of the data is not realised. Here we adapt a least-cost-path solver and specially
tailored cost-functions to rapidly extract and measure structural features from point cloud and raster datasets. We
implement the method in the geographic information system QGIS and the point cloud and mesh processing
software CloudCompare. Using these implementations, the method can be applied to a variety of three-
dimensional (3D) and two-dimensional (2D) datasets including high-resolution aerial imagery, virtual outcrop
15
models, digital elevation models (DEMs) and geophysical grids.
We demonstrate the algorithm with four diverse applications, where we extract: (1) joint and contact patterns in
high-resolution orthophotographs; (2) fracture patterns in a dense 3D point cloud; (3) earthquake surface ruptures
of the Greendale Fault associated with the Mw7.1 Darfield earthquake (New Zealand) from high-resolution light
detection and ranging (LiDAR) data, and; (4) oceanic fracture zones from bathymetric data of the North Atlantic.
20
The approach improves the objectivity and consistency of the interpretation process while retaining expert-
guidance, and achieves significant improvements (35-65%) in digitisation time compared to traditional methods.
Furthermore, it opens up new possibilities for data synthesis and can quantify the agreement between datasets and
an interpretation.
1 Introduction
25
Remote sensing datasets are commonly used in the earth sciences to interpret the morphology, location, timing
and orientation of geological features. These data types, now routinely delivered by satellite, aerial and UAV
platforms, have advanced to the point where they are widely available at high-resolution, and in some instances
frequently updated. This proliferation of data has led to a situation where it is now no longer practical to use
manual methods to extract geological information, meaning that the full geological value of high-quality datasets
30
is often not extracted.
For example, high and ultra-high resolution (cm to mm) photorealistic reconstructions of geological outcrops
(“virtual outcrop models”) are becoming widely available (Bemis et al., 2014; De Paor, 2016), typically acquired
using either laser scanning technology (cf. Buckley et al., 2008) or photogrammetric workflows (cf. Bemis et al.,
2014). It is feasible to use these techniques to capture areas of several square kilometres at mm-to-cm resolution
35
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
2
using off-the-shelf and easy to UAV technology (e.g., Vollgger and Cruden, 2016; Cruden et al., 2016), providing
for the first time an objective method for rapidly collecting detailed 3D information on geological structures.
There has recently been significant effort to develop automatic or computer-assisted methods for digitising
structural data, in particular from orthorectified photographs or image sequences (Seers and Hodgetts, 2016;
Vasuki et al., 2014; Jones et al., 2009). Achieving satisfactory automated digitisation is challenging for the
5
mapping of geological structures due to intrinsic variables such as geometry, soft-linkage and segmentation over
multiple scales, as well as extrinsic variables such as natural variations in colour, shadows, glare, and/or
incomplete geological exposure. Due to this complexity, fully automatic methods often require significant manual
adjustment and vetting to remove false positives while retaining real geological features (Vasuki et al., 2014;
Seers and Hodgetts, 2016).
10
In this paper, we first review existing approaches to the mapping of geological structures and contacts from digital
data, and then describe a novel least-cost path method that can “follow” structure traces and lithological contacts
between user-defined control points in both 2D gridded datasets (photographs, geophysical imagery etc.) and
dense 3D point clouds (virtual outcrop models). We then describe its implementation in two widely used software
packages (QGIS and CloudCompare), and introduce four applications demonstrating the efficacy of the method
15
for mapping outcrops, earthquake surface ruptures and oceanic fracture zones.
2 Existing methods
Many automated methods have been developed to extract linear features in the geosciences (e.g., Tzong-Dar and
Lee, 2007; Jinfei and Howarth, 1990). These use computer vision algorithms for edge and lineament detection
and, while often successful in ideal situations, require substantial fine-tuning to achieve optimal performance on
20
real-world data. They also have a tendency to detect many false positives related to non-geological features such
as shadows, roads or vegetation (Vasuki et al., 2014). Hence, even fully automated methods currently require
significant manual effort to remove non-geological features while ensuring features of interest are correctly
detected.
To circumvent these difficulties, several methods have been developed which remain user-driven but also use
25
computational power to optimise the interpretation process and improve objectivity and consistency. Vasuki et al.
(2014), for example, use an edge detection algorithm (phase congruency; cf. Kovesi, 1999) on orthophotographs
to optimise manually defined fracture traces and contacts. This allows the user to quickly define the approximate
locations of interesting features and then automatically refine them, speeding up the digitisation process
significantly while avoiding problems associated with false positives. Similar computer-assisted approaches have
30
also been applied to improve the interpretation of faults in regional magnetic surveys (Holden et al., 2016) and
oceanic fracture zones in global gravity datasets (Wessel et al., 2015).
In many situations, the 3D orientations of detected features are of interest. This is typically calculated using a
digital elevation model to add height information to each trace and then computing a best-fit plane (e.g., Dering
et al., 2016; Jaboyedoff et al., 2009; Banerjee and Mitra, 2005). While this method works well in simple
35
topography, it is inherently limited to 2.5 dimensions (2.5D), causing problems when features crosscut steep or
overhanging topography (Pavlis and Mason, 2017). For this reason, direct analysis of 3D point cloud data is
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
3
preferable over methods that are limited to 2.5D. Unfortunately, the unstructured nature of 3D point data mean
that methods for trace detection in raster data, such as those described above, cannot be easily applied.
A number of automatic methods for analysing point-cloud data have been proposed. These use clustering or plane-
fitting algorithms to automatically segment and extract joint or bedding faces (facets) exposed on the surface of
the outcrop, with reasonable success (e.g., Dewez et al., 2016; Lato and Vöge, 2012; García-Sellés et al., 2011).
5
However, structural surfaces are not always directly exposed, and instead intersect the outcrop surface to form
linear features referred to as structural traces. These traces cannot be detected using facet-based techniques, and
require a different approach.
Seers and Hodgetts (2016) demonstrate one such approach, automatically extracting 3D structural traces by
applying image-based edge detection techniques (phase congruency) to a set of images and then projecting the
10
identified traces into 3D, using depth information derived from photogrammetric reconstructions or associated
laser scan data. This approach uses multiple images to overcome issues associated with out-of-plane geometry,
however as with other fully automated methods, a variety of parameters and thresholds require careful calibration
and the results must be manually vetted to remove false positives.
3 Method: A least-cost path approach to digital mapping
15
3.1. Theory
The approach presented here couples algorithms for solving least-cost path problems with both general and use-
case specific cost functions to capture structural features in both point-cloud and raster datasets. Least-cost path
algorithms have previously been used to detect linear features in a variety of image data and have proven robust
even when signal-to-noise ratios are very low (Sun and Pallottino, 2003; Vincent, 1998; Buckley and Yang, 1997).
20
Conceptually the algorithm can be divided into two steps, although for performance reasons our implementation
performs these simultaneously. In the first step, data points (points in a point cloud or pixels in an image) are
linked with their nearest neighbours, using a spherical search radius slightly larger than the dataset resolution, to
produce a neighbourhood network (Fig. 1a). The costs of moving along links in this network (hereafter referred
to as “edges”) are then calculated, using a cost function designed to promote movement along structural or
25
lithological traces and inhibit movement across them (Fig. 1b).
In the second step, an optimised version of Dijkstra’s algorithm (Dijkstra, 1959) is used to derive the least-cost
path between user-defined control points, providing the estimated trace (Fig. 1c). Djikstra’s algorithm, in essence,
progressively “grows” least-cost paths from the start point until the end is found. We optimise this by requiring
paths to move closer to the end at each step, eliminating tortuous geometries that tend not to be geologically
30
feasible. Once a trace has been estimated, manual adjustments can be easily applied by adding intermediate
waypoints and recalculating the relevant least-cost paths.
The critical component in this approach is the cost function. A well-designed cost function produces low values
for edges following structure or contact traces, and high values for edges outside or crosscutting traces. Our
optimised implementation of Dijkstra’s algorithm then follows edges with the lowest cost-values in order to map
35
out the feature of interest. We have designed and implemented five simple cost functions that give reasonable
results for different structure and data types (Appendix 1). Conveniently, simple cost functions such as point/pixel
brightness or local colour gradient work well on most geological datasets; the examples presented below all map
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
4
a single scalar attribute in the dataset directly to cost (point/pixel brightness for Study 1 and 2, topographic slope
for Study 3 and bathymetric depth and vertical gravity gradient for Study 4).
3.2. Implementation
The above methodology has been implemented as plugins for Cloud Compare (Girardeau-Montaut, 2015) and
QGIS (QGis, 2011), both of which are cross-platform, open-source and widely used software packages for
5
geospatial analysis. Our CloudCompare plugin (Compass) works on point clouds, while the QGIS implementation
(GeoTrace) works on raster data. Compass is bundled with the default CloudCompare distribution (since version
2.9), and the source code is freely available at https://github.com/CloudCompare/CloudCompare. Similarly,
GeoTrace can be found on the QGIS plugin repository (https://plugins.qgis.org/plugins/), and the source code
downloaded from https://github.com/lachlangrose/GeoTrace. Complete documentation for the plugins is found at
10
the CloudCompare wiki (http://www.cloudcompare.org/doc/wiki/) and on the GeoTrace GitHub page.
In addition to our method for rapidly extracting structural traces, a variety of other functionality has been
implemented, including tools for measuring surface orientations, lineations and true thicknesses in the Cloud
Compare plugin, and DEM based plane fitting and orientation analysis in the QGIS plugin.
4 Case Studies
15
To demonstrate the capability of the computer-assisted trace detection approach described above, we present the
results of four case studies. These studies highlight the versatility of our method and its increased efficiency as
compared to established manual methods.
The first case study involves the interpretation of joint sets in two 10 × 10 m areas from a ~1 cm resolution
orthophotograph of a wavecut platform at Bingie Bingie Point, New South Wales, Australia. The outcrop contains
20
several Cretaceous to Paleogene dykes intruding Devonian plutonic diorites and tonalities and crosscut by a series
of complex joint sets. The orthophotograph was generated by applying a Structure from Motion-Multi-View
Stereo (SfM-MVS) workflow (Cruden et al., 2016) to digital photographs captured from a DJI S800 Evo multi-
rotor UAV fitted with a 24.3-megapixel Sony Nex-7 camera and 16mm F2.8 prime lens. The two 10 × 10 m areas
(Fig. 2a, b) were selected from the survey as they contain well exposed dykes and joint sets as well as common
25
confounding effects such as shadows and puddles. For demonstration purposes the selected areas are relatively
small, but the workflow is equally applicable to much larger outcrops.
Our second case study focuses on the extraction of 3D joint traces and orientations, which are interpreted directly
on a dense 3D point cloud. The Cape Woolamai sea stacks, located approximately 115 km southeast of Melbourne
on Phillip Island, have formed by erosion of the coarse-to medium-grained Cape Woolamai granite, which
30
intruded Silurian to Lower Devonian meta-turbidites during or slightly after the mid-Devonian Tabberabberan
Orogeny (a widespread episode of deformation and plutonism across Victoria; Gray, 1997; Richards and
Singleton, 1981). Several sets of systematic and non-systematic joints crosscut this granite, likely related to the
cooling of the intrusion, subsequent deformation and recent unloading.
For this study, a DJI Inspire 1 multi-rotor UAV and a Zenmuse X3 camera were used to capture aerial photographs,
35
which were subsequently processed using a SfM-MVS workflow. The resulting 3D model is 45 × 40 × 25 m in
size and comprises ~2 million points, with a ground sampling distance of ~2.5 cm/pixel. The topographic
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
5
complexity of this outcrop allows for accurate orientation measurement, but makes interpretation from 2.5D
datasets (orthophotograph + DEM) impractical (Fig. 3a, b).
For the third case study, surface ruptures that formed along the Greendale Fault after the 2010 Mw7.1 Darfield
earthquake are extracted from a 1 m resolution LiDAR derived DEM. The data was collected a few days after the
earthquake and was used, along with a variety of other data, to measure the surface displacement resulting from
5
the earthquake and to interpret the kinematics of the Greendale Fault (Duffy et al., 2012).
Finally, for our last case study we interpret oceanic fracture zones in the North Atlantic from 30 arc-second
bathymetry (Weatherall et al., 2015) and vertical gravity gradient data (Sandwell et al., 2014). From their inception
at mid-ocean ridges, fracture zones can be used to constrain plate motion vectors and are widely used in tectonic
reconstruction (Williams et al., 2016; Sandwell et al., 2014). Both these datasets provide an opportunity to test
10
our method on global-scale geophysical data.
5 Results
In each of the case studies described above the data have been interpreted twice, once using our computer-assisted
method and once using manual workflows in QGIS (Study 1, 3 and 4) or CloudCompare (Study 2). The time
required to extract comparable amounts of structural data was significantly reduced using the computer-assisted
15
method (Table 1). This efficiency increase was especially pronounced (61%) for the point cloud example, as
manual methods for digitising linear features on 3D point clouds are particularly time consuming.
The following four subsections compare and contrast the results of both manual and assisted interpretations in
more detail.
5.1. Bingie Bingie Point
20
Both areas (Fig. 2a, b) of the Bingie Bingie Point orthophotographs contain joints over a range of scales and in a
variety of host rocks, as well as features that make automated interpretation challenging such as water, shadows
and debris-filled joints. Fracture and contact traces were digitised manually in QGIS (Fig. 2c, d), and with the
GeoTrace implementation of our assisted method (Fig. 2e, f). For the assisted interpretation, different cost
functions were used to pick the fractures and the dyke contacts. Fractures in the orthophotographs are clearly
25
darker than their surroundings, so a greyscale version of the orthophotograph (easily calculated using GeoTrace)
was used to define the shortest-path cost function during fracture digitisation. Dyke contacts were mapped using
a cost function derived from the inverse of the local brightness gradient (high gradient = low cost). This was
achieved by applying a Sobel filter (essentially a local gradient operator) to the previously mentioned greyscale
image, using scikit-image functionality (van der Walt et al., 2014) integrated into GeoTrace.
30
The results are visually similar to the manually derived reference interpretation (Fig. 2c-f). Closest-point
differences between the manual and assisted interpretations show that the majority of traces (78% in Area 1 and
70% in Area 2) match to within 2 pixels (≈2 cm), smaller than the ambiguity of the dataset.
5.2. Cape Woolamai
Joints in the Cape Woolamai virtual outcrop model were interpreted in 3D using CloudCompare, first with the
35
manual “draw polyline” tool and then using the Compass implementation of our method. The complex topography
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
6
of the sea-stacks makes 2.5D analysis inappropriate (Fig. 3a, b). As in the previous example, cost was defined by
point brightness as fractures are defined by their darker colour.
In total, 146 joint traces were interpreted manually over ~3 hours, while 114 joint traces were digitised using the
Compass plugin in less than an hour (Table 1). Joint orientations were estimated by calculating the least-squares
plane-of-best-fit for each trace. The ratio between the second and third eigenvectors of each trace was then used
5
to reject arbitrary planes resulting from sub-linear traces, using a planarity threshold of 0.75 (cf. Thiele et al., 2015
for a more detailed description of this method). Compass does this in real-time during the digitisation processes,
while orientation estimates from the manually digitised dataset were calculated as a post-processing step. The
manual and computer-assisted methods resulted in 133 and 91 orientation estimates respectively.
Both sets of interpreted traces and associated orientation estimates appear to be broadly consistent for each method
10
(Fig. 3c-f). Significantly, orientation estimates from the computer assisted method form more-pronounced clusters
than equivalents estimated using the manually digitised traces. Although far from conclusive, this indicates that
the computer assisted approach improves the consistency and precision of the orientation estimates.
5.3. Greendale Fault
Surface ruptures of the Greendale Fault form a series of en échelon fault-scarps visible in the LiDAR dataset (Fig.
15
4a). Our shortest-path method can be used to pick the fault scarps using a cost function where slope maps inversely
with cost. This was achieved by calculating a slope raster using the QGIS DEM (Terrain models) tool and inverting
it using GeoTrace.
As in the previous examples the assisted interpretation achieved very similar results to a manual interpretation, in
about half the time. Closest-point difference calculations between the manual and assisted traces also show the
20
two sets of interpretation are consistently within ~12 pixels (~2 m).
5.4. Oceanic Fracture Zones
Oceanic fracture zones in the North Atlantic were digitised in GeoTrace using bathymetric depth to define trace
cost. Comparison with an interpretation that was digitised manually shows similar accuracies to the previous case
studies, with the majority of traces within ~2 pixels, as well as an improvement of 36% in per-trace digitisation
25
time (Fig. 5).
Additionally, we used the start and end points of oceanic fracture zones described in Matthews et al. (2011), which
are based on a 2009 gravity gradient compilation (Sandwell and Smith, 2009), to constrain an otherwise unguided
GeoTrace interpretation of an updated vertical gravity gradient dataset (Sandwell et al., 2014). This was achieved
using the vertical gravity gradient directly as the cost function, such that traces follow areas of low vertical
30
gradient, and then solving the shortest-path between the Matthews et al. (2011) start and end points.
The results (Fig. 5d) again highlight the tool’s general accuracy, with 65% of traces falling within 2 pixels of the
Matthews et al. (2011) interpretation and 79% within 5 pixels. Most errors occurred in areas of closely spaced
fracture zones, where the computed shortest-path for many fracture zones would “detour” through adjacent low-
cost features (Fig. 6). A small number of additional control points along these traces resolve this issue (Fig. 6c)
35
by forcing the computed path to stay in the local cost-minima (the correct fracture zone), rather than taking
advantage of larger adjacent minima.
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
7
6. Discussion
The four case studies presented above highlight applications of the least-cost-path method to the interpretation of
high-resolution aerial orthophotographs, 3D point clouds, LiDAR DEM and bathymetric data - all datasets
commonly used in the earth sciences to interpret and characterise geological features. We discuss here three
aspects of our least-cost-path approach: it improves objectivity and reproducibility, allows automatic refinement
5
if better data becomes available, and unlike fully-automated workflows, operates in co-operation with expert
guidance.
Firstly, the approach is more objective than manual digitisation. Although not as objective as fully-automated
methods (the location of the trace start and end are interpreted), most of the length of each trace is determined
algorithmically, and hence will consistently locate in the same spot. Indeed, as demonstrated in Figure 7, the
10
calculated shortest path varies only slightly when control points are interpreted at different locations. Results from
Study 2 indicate that this improved consistency might increase the precision of the derived orientation estimates
(Fig. 3e, f). Furthermore, each control point can easily be stored, providing a record of the locations at which
interpretive decisions were made.
Similarly to the method outlined by Wessel et al. (2015) for extracting oceanic fracture zones, these control points
15
can also be reused to generate an updated interpretation if higher-resolution or more accurate information become
available. This possibility is demonstrated in Study 4, where published oceanic fracture zones were reconstructed
automatically using an updated underlying dataset and the start and end points of a previous interpretation
(Matthews et al., 2011). Although some quality control is required after such an operation, the digitisation process
no longer needs to be completely repeated, and interpretations can be rapidly updated as datasets evolve.
20
Where multiple datasets are available, the similarity and total cost of paths reconstructed using different datasets
can be used to quantitatively assess the degree to which different datasets support an interpretation. It is common
in the geosciences to bring interpretations from multiple types of data into a single synthesis (e.g., Seton et al.,
2016; Blaikie et al., 2017), especially when using geophysical datasets such as gravity and magnetics. Limiting
factors during such data synthesis include both the time required and highly subjective nature of multi-data type
25
interpretations, so a method for rapidly quantifying the extent to which different datasets support an interpretation
serves as an important addition. Similarly, sensitivity analyses could be performed by randomly moving control
points and measuring the response of the traces to quantify the robustness of the interpretation to uncertainty
(similar to Fig. 7).
The time it takes for users to manually interpret datasets using GeoTrace or Compass will vary significantly
30
between users, and the purpose of this study was not to comprehensively measure the efficiency of our approach.
Nevertheless, in each of the case studies, our initial assessment indicates that computer-assisted interpretation
required ~35-66% less user effort, as measured by both average time and mouse-clicks per structure trace, when
compared to manual methods (Table 1). The resulting traces also appear to be comparable to manual traces in
each case (~±2 pixels), demonstrating that our method can be used to achieve equivalent results.
35
The Compass implementation of the technique produces especially impressive results, reducing interpretation
time in the Cape Woolamai example by 61%. This is pertinent given the rapid growth in both size and availability
of high-resolution point cloud data and the limited range of available tools for extracting structural data from
them. Significantly, the implementation of our least-cost-path method in Compass requires only local information,
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
8
such that calculation time scales with trace length and not dataset size. This means the tool can be used to interpret
arbitrarily large point clouds.
Finally, the computer-assisted philosophy behind our method keeps the expert in control of the entire digitisation
process, allowing data vetting and correction during digitisation. The approach ensures the expert becomes
familiar with the particular intricacies of each dataset, a key part of further data analysis and something not
5
possible using automated methods yet essential for the creative process of understanding and interpreting spatial
information.
7. Conclusions
We have described a least-cost-path based method for the computer-assisted digitisation of structural traces in
point cloud, image and raster datasets. The method enhances an expert’s ability to extract geological information
10
from the wide range of high-resolution data available to geoscientists while reducing the required time and effort.
In summary, the method:
Allows expert-guided interpretation in a way that seamlessly utilises computing power to significantly
optimise the interpretation process and improve objectivity and consistency.
Can be applied to both raster and point-cloud datasets. This is particularly significant in situations where
15
complex topography prevents a more conventional 2.5D raster based workflow.
Requires only local knowledge of a dataset, so that the total dataset size does not affect performance;
thereby allowing computer-assisted interpretation of exceedingly large datasets.
Is implemented as two freely available and open-source plugins for the widely used CloudCompare and
QGIS software packages.
20
Data Availability
Datasets used for the Bingie Bingie Point and Cape Woolamai case studies are freely available from
https://doi.org/10.4225/03/5981b31091af9. The bathymetric and vertical gravity gradient datasets used for the
oceanic fracture zone example can be downloaded from the University of California San Diego at
http://topex.ucsd.edu/grav_outreach/, while the Greendale Fault LiDAR dataset is available on request from the
25
authors of Duffy et al. (2012).
Author Contributions
ST and LG developed the methodology described in this study. ST implemented it in CloudCompare and LG
implemented it in QGIS. All of the authors contributed to the case studies and helped prepare the manuscript. The
authors declare that they have no conflict of interest.
30
Acknowledgements
The authors would like to gratefully acknowledge Daniel Girardeau-Montaut and other CloudCompare developers
for creating a fantastic software package and for their assistance creating the Compass plugin. ST was supported
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
9
by a Westpac Future Leaders Scholarship and Australian Postgraduate Award. LG was supported by an Australian
Postgraduate Award. AS was supported by a Monash University Faculty of Science Dean's International
Postgraduate Research Scholarship and an American Association of Petroleum Geologists Grants-in-Aid award.
References
Banerjee, S., and Mitra, S.: Foldthrust styles in the Absaroka thrust sheet, Caribou National Forest area, Idaho
5
Wyoming thrust belt, Journal of Structural Geology, 27, 51-65, 10.1016/j.jsg.2004.07.004, 2005.
Bemis, S. P., Micklethwaite, S., Turner, D., James, M. R., Akciz, S., Thiele, S. T., and Bangash, H. A.: Ground-
based and UAV-Based photogrammetry: A multi-scale, high-resolution mapping tool for structural geology and
paleoseismology, Journal of Structural Geology, 69, 163-178, 10.1016/j.jsg.2014.10.007, 2014.
Blaikie, T. N., Betts, P. G., Armit, R. J., and Ailleres, L.: The ca. 17401710Ma Leichhardt Event: Inversion of a
10
continental rift and revision of the tectonic evolution of the North Australian Craton, Precambrian Research, 292,
75-92, http://dx.doi.org/10.1016/j.precamres.2017.02.003, 2017.
Buckley, M., and Yang, J.: Regularised shortest-path extraction, Pattern Recognition Letters, 18, 621-629,
10.1016/s0167-8655(97)00076-7, 1997.
Buckley, S. J., Howell, J. A., Enge, H. D., and Kurz, T. H.: Terrestrial laser scanning in geology: data acquisition,
15
processing and accuracy considerations, Journal of the Geological Society, 165, 625-638, 10.1144/0016-
76492007-100, 2008.
Cruden, A., Vollgger, S., Dering, G., and Micklethwaite, S.: High Spatial Resolution Mapping of Dykes Using
Unmanned Aerial Vehicle (UAV) Photogrammetry: New Insights On Emplacement Processes, Acta Geologica
Sinica (English Edition), 90, 52-53, 10.1111/1755-6724.12883, 2016.
20
De Paor, D. G.: Virtual Rocks, GSA Today, 4-11, 10.1130/gsatg257a.1, 2016.
Dering, G., Micklethwaite, S., Barnes, S. J., Fiorentini, M., Cruden, A., and Tohver, E.: An Elevated Perspective:
Dyke-Related Fracture Networks Analysed with Uav Photogrammetry, Acta Geologica Sinica - English Edition,
90, 54-55, 10.1111/1755-6724.12884, 2016.
Dewez, T. J. B., Girardeau-Montaut, D., Allanic, C., and Rohmer, J.: Facets : A CloudCompare plugin to extract
25
geological planes from unstructured 3D point clouds ISPRS - International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Sciences, XLI-B5, 799-804, 10.5194/isprsarchives-xli-b5-799-2016,
2016.
Dijkstra, E. W.: A note on two problems in connexion with graphs, Numerische Mathematik, 1, 269-271,
10.1007/bf01386390, 1959.
30
Duffy, B., Quigley, M., Barrell, D. J. A., Van Dissen, R., Stahl, T., Leprince, S., McInnes, C., and Bilderback, E.:
Fault kinematics and surface deformation across a releasing bend during the 2010 MW 7.1 Darfield, New Zealand,
earthquake revealed by differential LiDAR and cadastral surveying, Geological Society of America Bulletin, 125,
420-431, 10.1130/b30753.1, 2012.
García-Sellés, D., Falivene, O., Arbués, P., Gratacos, O., Tavani, S., and Muñoz, J. A.: Supervised identification
35
and reconstruction of near-planar geological surfaces from terrestrial laser scanning, Computers & Geosciences,
37, 1584-1594, 10.1016/j.cageo.2011.03.007, 2011.
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
10
Gray, D. R.: Tectonics of the southeastern Australian Lachlan Fold Belt: structural and thermal aspects,
Geological Society, London, Special Publications, 121, 149-177, 10.1144/gsl.sp.1997.121.01.07, 1997.
Holden, E.-J., Wong, J. C., Wedge, D., Martis, M., Lindsay, M., and Gessner, K.: Improving assessment of
geological structure interpretation of magnetic data: An advanced data analytics approach, Computers &
Geosciences, 87, 101-111, 10.1016/j.cageo.2015.11.010, 2016.
5
Jaboyedoff, M., Couture, R., and Locat, P.: Structural analysis of Turtle Mountain (Alberta) using digital elevation
model: Toward a progressive failure, Geomorphology, 103, 5-16, 10.1016/j.geomorph.2008.04.012, 2009.
Jinfei, W., and Howarth, P. J.: Use of the Hough transform in automated lineament, IEEE Transactions on
Geoscience and Remote Sensing, 28, 561-567, 10.1109/tgrs.1990.572949, 1990.
Jones, R. R., McCaffrey, K. J. W., Clegg, P., Wilson, R. W., Holliman, N. S., Holdsworth, R. E., Imber, J., and
10
Waggott, S.: Integration of regional to outcrop digital data: 3D visualisation of multi-scale geological models,
Computers & Geosciences, 35, 4-18, http://dx.doi.org/10.1016/j.cageo.2007.09.007, 2009.
Kovesi, P.: Image features from phase congruency, Videre: Journal of computer vision research, 1.3, 1-26, 1999.
Lato, M. J., and Vöge, M.: Automated mapping of rock discontinuities in 3D lidar and photogrammetry models,
International Journal of Rock Mechanics and Mining Sciences, 54, 150-158, 10.1016/j.ijrmms.2012.06.003, 2012.
15
Matthews, K. J., Müller, R. D., Wessel, P., and Whittaker, J. M.: The tectonic fabric of the ocean basins, Journal
of Geophysical Research, 116, 10.1029/2011jb008413, 2011.
Pavlis, T. L., and Mason, K. A.: The New World of 3D Geologic Mapping, GSA Today, 10.1130/gsatg313a.1,
2017.
Richards, J. R., and Singleton, O. P.: Palaeozoic Victoria, Australia: Igneous rocks, ages and their interpretation,
20
Journal of the Geological Society of Australia, 28, 395-421, 10.1080/00167618108729178, 1981.
Sandwell, D. T., and Smith, W. H. F.: Global marine gravity from retracked Geosat and ERS-1 altimetry: Ridge
segmentation versus spreading rate, Journal of Geophysical Research, 114, 10.1029/2008jb006008, 2009.
Sandwell, D. T., Muller, R. D., Smith, W. H. F., Garcia, E., and Francis, R.: New global marine gravity model
from CryoSat-2 and Jason-1 reveals buried tectonic structure, Science, 346, 65-67, 10.1126/science.1258213,
25
2014.
Seers, T. D., and Hodgetts, D.: Extraction of three-dimensional fracture trace maps from calibrated image
sequences, Geosphere, 12, 1323-1340, 2016.
Seton, M., Mortimer, N., Williams, S., Quilty, P., Gans, P., Meffre, S., Micklethwaite, S., Zahirovic, S., Moore,
J., and Matthews, K. J.: Melanesian back-arc basin and arc development: Constraints from the eastern Coral Sea,
30
Gondwana Research, 39, 77-95, https://doi.org/10.1016/j.gr.2016.06.011, 2016.
Sun, C., and Pallottino, S.: Circular shortest path in images, Pattern Recognition, 36, 709-719, 10.1016/s0031-
3203(02)00085-7, 2003.
Thiele, S. T., Micklethwaite, S., Bourke, P., Verrall, M., and Kovesi, P.: Insights into the mechanics of en-échelon
sigmoidal vein formation using ultra-high resolution photogrammetry and computed tomography, Journal of
35
Structural Geology, 77, 27-44, 10.1016/j.jsg.2015.05.006, 2015.
Tzong-Dar, W., and Lee, M. T.: Geological lineament and shoreline detection in SAR images, 2007 IEEE
International Geoscience and Remote Sensing Symposium, 2007.
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
11
van der Walt, S., Schönberger, J. L., Nunez-Iglesias, J., Boulogne, F., Warner, J. D., Yager, N., Gouillart, E., and
Yu, T.: scikit-image: image processing in Python, PeerJ, 2, e453, 10.7717/peerj.453, 2014.
Vasuki, Y., Holden, E.-J., Kovesi, P., and Micklethwaite, S.: Semi-automatic mapping of geological Structures
using UAV-based photogrammetric data: An image analysis approach, Computers & Geosciences, 69, 22-32,
10.1016/j.cageo.2014.04.012, 2014.
5
Vincent, L.: Minimal path algorithms for the robust detection of linear features in gray images, Computational
Imaging and Vision, 12, 331-338, 1998.
Vollgger, S. A., and Cruden, A. R.: Mapping folds and fractures in basement and cover rocks using UAV
photogrammetry, Cape Liptrap and Cape Paterson, Victoria, Australia, Journal of Structural Geology, 85, 168-
187, https://doi.org/10.1016/j.jsg.2016.02.012, 2016.
10
Weatherall, P., Marks, K. M., Jakobsson, M., Schmitt, T., Tani, S., Arndt, J. E., Rovere, M., Chayes, D., Ferrini,
V., and Wigley, R.: A new digital bathymetric model of the world's oceans, Earth and Space Science, 2, 331-345,
10.1002/2015ea000107, 2015.
Wessel, P., Matthews, K. J., Müller, R. D., Mazzoni, A., Whittaker, J. M., Myhill, R., and Chandler, M. T.:
Semiautomatic fracture zone tracking, Geochemistry, Geophysics, Geosystems, 16, 2462-2472,
15
10.1002/2015gc005853, 2015.
Williams, S. E., Flament, N., and Müller, R. D.: Alignment between seafloor spreading directions and absolute
plate motions through time, Geophysical Research Letters, 43, 1472-1480, 10.1002/2015gl067155, 2016.
20
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
12
Appendix 1
We outline five simple cost functions that give reasonable results for different structure types. Each function is
designed to give values between 0 and 1, allowing combinations of functions to be used (by summation), and to
work on both unstructured datasets (i.e. point clouds) and structured datasets (images). Hence, we do not present
any cost functions that rely on commonly used image processing techniques such as edge enhancement, although
5
these could be easily incorporated for raster datasets. These functions are implemented directly in the Compass
plugin, while simple QGIS functionality can be used to apply them to raster data for use with GeoTrace.
Colour Brightness
The brightness of an edge’s end colour (eRGB) can be mapped directly to edge cost (the brightness of an edges start
colour will be incorporated into the previous edge in the path). Despite its simplicity, this function (Eq. 1) is
10
surprisingly effective at picking fracture traces, which are typically darker than their surroundings due to
shadowing. Similarly, bright traces such as thin quartz or calcite veins can be identified using the opposite of this
cost function (Eq. 2). Note that the division by 3 ensures that the function maps to the 0 1 range (assuming red,
green and blue values also range from 0 to 1).
15
3e_B+e_G+e_R
=cost
(Eq. 1)
3e_B+e_G+e_R
- 1 =cost
(Eq. 2)
Colour Similarity
A similar cost function, based on colour similarity rather than brightness alone, is useful in more generic situations
20
where traces have a distinctive colour but are not necessarily darker or lighter than their surroundings. This
function (Eq. 3) considers an edge to be low cost if: (1) the start and end colours are similar, and; (2) the start and
end colours are similar to the colour of the start and end of the trace (BRGB and ERGB), minimizing along-path
gradient and maximizing similarity with the trace start and end points. This function works well when traces have
a specific colour, such as for cemented joints, though it is comparatively slow compared to the brightness-based
25
functions described above due to the large number of square roots. Similarly to the previous equations, the factors
of 3 ensure that the function maps to the 0 1 range.
34
2
1
3
2
1
=cost RGBRGBRGBRGBRGBRGBRGBRGBRGBRGB EeBeEsBses
(Eq. 3)
Gradient
The previous cost functions are useful for identifying discrete structural traces such as faults, joints or thin veins,
30
but will not be sensitive to lithological contacts. Lithological contacts are typically defined by changes in colour,
and hence we base a cost function around colour gradient to identify them. This function (Eq. 4) evaluates the
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
13
gradient G[N] of the magnitude of the colour vectors across the start and end neighbourhoods Nstart and Nend. To
calculate the gradient for point cloud data, we use a simple method that calculates the average distance-weighted
point-to-point gradient for each neighbourhood. More complex methods would highlight contacts better, but at a
computational cost. For raster data, we implement a Sobel filter to achieve equivalent results.
An upper limit (l) is applied to the gradient in order to maintain a cost value between 0 and 1. A reasonable value
5
for this limit can be approximated by dividing the maximum change in colour magnitude (3 ) by the average
distance between data points. This cost function can also be improved by log-transforming it to increase the
importance of gradients resulting from more subtle features.
 
 
l,Gmin
- 1 =cost start lNGN end
(Eq. 4)
10
Curvature
In some situations, resolution is high enough that structural traces, fractures in particular, appear as topographic
ridges or valleys. Hence, we include a final cost function (Eq. 5) which considers points with a high curvature as
low cost, allowing paths to ‘follow’ ridges and valleys, where C[N] calculates the mean curvature of a point or
pixel neighbourhood N, and l is an arbitrarily large upper limit (that allows the log-curvature to scale from 0 to
15
1). Note that calculating the mean curvature of a neighbourhood is computationally expensive, so this cost function
performs significantly slower than the previously described ones unless curvature is pre-computed.
 
 
l,Clogmin
- 1 =cost end lN
(Eq. 5)
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
14
Table 1. Manual vs computer-assisted digitisation for the different study areas. Percentage improvements are
calculated by comparing the average time and mouse clicks per digitised trace. Each case study shows a clear reduction
in digitisation time, especially for the 3D datasets where manual interpretation can be especially tedious.
Method
Number
of traces
Improvement
%
Number of
mouse clicks
Improvement
%
Study 1: Bingie Bingie
Area 1
Manual
0:54
270
-
2253
-
Assisted
0:37
283
35%
917
61%
Area 2
Manual
0:57
338
-
2509
-
Assisted
0:35
383
46%
1122
61%
Study 2: Cape Woolamai
Manual
3:04
146
-
6026
-
Assisted
0:56
114
61%
1703
64%
Study 3: Greendale Fault
Manual
0:18
74
-
1039
-
Assisted
0:07
93
51%
282
66%
Study 4: Oceanic Fracture Zones
Manual
1:17
432
-
5731
-
Assisted
0:35
310
36%
1265
69%
5
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
15
Figure 1. Schematic representation of the least-cost path approach to trace detection for point cloud (top) and raster
(bottom) data. Points/pixels on the structural trace have a lower brightness in this example (a), so a brightness-based
cost function will result in low-cost edges between adjacent points/pixels that both fall on the structure trace (b). A
least-cost path calculation (c) then provides an estimate of the structure trace.
5
Figure 2. The two 10 × 10 m orthophotographs (a, b) interpreted in Study 1. Fracture traces were digitised manually
(c, d) and with our assisted method (e, f). Closest-point distances between the assisted and manual interpretations are
also shown (g, h). Note the tails of these distributions have been clipped to 5 cm, as some assisted traces did not have
10
manual equivalents, and hence gave incorrectly large closest-point differences. Small crosses in (e) and (f) represent
the control points that were digitised by the user to constrain the shortest path algorithm.
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
16
Figure 3. Orthophotograph of the Cape Woolamai sea stacks (a) and oblique view of the equivalent dense point cloud
(b). A 2.5D analysis conducted using the orthophotograph would significantly under sample the moderately to shallowly
dipping joint sets which are clearly visible on sub-vertical exposures in (b). Hence fractures were digitised in 3D, both
manually (c) and using the computer assisted approach (d). Equal-area lower hemisphere stereographic projections of
5
poles to joint orientations estimated from each of these interpretations (e-f) show that both methods produce similar
results. Poles from the computer-assisted dataset cluster more tightly (maximum density = 14.7%) than the manually
interpreted dataset (maximum density = 8.1%), indicating that the computer-assisted approach results in more
consistent orientation estimates.
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
17
Figure 4. LiDAR dataset illuminated from the NW showing surface ruptures of a section of the Greendale Fault, New
Zealand, collected shortly after the Mw7.1 Darfield earthquake (a). Traces interpreted manually (b) and using the
GeoTrace implementation of our least-cost-path method (c) are essentially equivalent (d). Control points for the assisted
interpretation are shown as small crosses.
5
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
18
Figure 5. Manual (a) and assisted (b) interpretations of oceanic fracture zones in the north Atlantic. Fracture zones
interpreted manually (c) from vertical gravity gradient by Matthews et al. (2011) and reconstructed in GeoTrace using
the start and end points only (d) are also shown. Red and blue colours in (c) and (d) show areas of high and low vertical
gravity gradient, respectively. As in the previous case studies, most equivalent manual and assisted traces fall within 2
5
pixels (e), though differences of up to 80 pixels occur in the reconstructed dataset (d).
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
19
Figure 6. Example of a larger cost minima (a) causing the incorrect reconstruction (b) of an oceanic fracture zone. In
this case the trace can be corrected by adding a single additional control point midway along the fracture zone (c).
5
Figure 7. Fracture (a) from the Bingie Bingie Point dataset showing that the majority of the resulting trace (b)
consistently follows the same path despite variation of the location of the control points.
Solid Earth Discuss., https://doi.org/10.5194/se-2017-83
Manuscript under review for journal Solid Earth
Discussion started: 15 August 2017
c
Author(s) 2017. CC BY 4.0 License.
... In the geosciences, UAV-based SfM-MVS (hereafter UAV-SfM) techniques have primarily been used for generating and analyzing two-dimensional (2-D) and two-and-a-half-dimensional (2.5-D) data sets, such as orthomosaic images and/or digital surface models (DSMs; Carrivick et al., 2016). Although these 2-D and 2.5-D data sets may be suitable for measurement and interpretation of geologic features along planar surfaces, they are susceptible to compression, distortion, and overgeneralization of details exposed along planes nonnormal to image acquisition (Bellian et al., 2005;Pavlis and Mason, 2017;Thiele et al., 2017). Geologic features are commonly exposed along steep slopes and are particularly susceptible to these effects; therefore, further consideration of data collection and visualization is necessary. ...
... UAV-SfM applications in geology have primarily used 2-D and 2.5-D methods for analysis (e.g., Bemis et al., 2014;Johnson et al., 2014;Vasuki et al., 2014;Chen et al., 2015;Zahm et al., 2016;Chesley et al., 2017;Thiele et al., 2017). ...
Article
Full-text available
Fluvial deposits are highly heterogeneous and inherently challenging to map in outcrop due to a combination of lateral and vertical variability along with a lack of continuous exposure. Heavily incised landscapes, such as badlands, reveal continuous three-dimensional (3-D) outcrops that are ideal for constraining the geometry of fluvial deposits and enabling reconstruction of channel morphology through time and space. However, these complex 3-D landscapes also create challenges for conventional field mapping techniques, which offer limited spatial resolution, coverage, and/or lateral contiguity of measurements. To address these limitations, we examined an emerging technique using images acquired from a small unmanned aerial vehicle (UAV) and structure-from-motion (SfM) photogrammetric processing to generate a 3-D digital outcrop model (DOM). We applied the UAV-SfM technique to develop a DOM of an Upper Cretaceous channel-belt sequence exposed within a 0.52 km2 area of Dinosaur Provincial Park (southeastern Alberta, Canada). Using the 3-D DOM, we delineated the lower contact of the channel-belt sequence, created digital sedimentary logs, and estimated facies with similar conviction to field-based estimations (±4.9%). Lateral accretion surfaces were also recognized and digitally traced within the DOM, enabling measurements of accretion direction (dip azimuth), which are nearly impossible to obtain accurately in the field. Overall, we found that measurements and observations derived from the UAV-SfM DOM were commensurate with conventional ground-based mapping techniques, but they had the added advantage of lateral continuity, which aided interpretation of stratigraphic surfaces and facies. This study suggests that UAV-SfM DOMs can complement traditional field-based methods by providing detailed 3-D views of topographically complex outcrop exposures spanning intermediate to large spatial extents.
... Vasuki et al. (2014),Seers and Hodgetts (2016),Thiele et al. (2017),Umili et al. (2013),Slob et al. (2005),Jaboyedoff et al. (2007),Gigli and Casagli (2011) et Assali et al. (2014) s'y sont intéressés pour proposer des solutions d'analyses semi-automatisées. Nous ne développons pas ces techniques dans ce manuscrit, car, pour notre projet de recherche, nous nous intéressons principalement aux surfaces de discontinuité dont les zones planaires sont visibles sur l'affleurement. ...
Thesis
Full-text available
Les lois récentes imposent de repérer l’amiante avant travaux. Ce cadre s’applique à l’amiante présent dans les matériaux de construction et dans les roches naturelles. L’objet du projet de thèse est de se placer à l’échelle locale d’un affleurement rocheux afin de proposer une carte 3D des zones amiantifères en exploitant les photographies des sites. Dans son contexte naturel, l’occurrence amiante est présente à la surface des fractures ayant des orientations réglées par l’histoire tectonique locale. Trois axes de recherche ont été suivis. Ils sont basés sur le traitement de nuages de points denses 3D obtenus par photogrammétrie.Le premier axe de recherche s’est focalisé sur la localisation spatiale et la caractérisation de l’orientation et de la fréquence des zones à forte densité de fractures d’un affleurement rocheux. Le deuxième s'est concentré sur l'optimisation des prises de vue pour restituer par photogrammétrie un affleurement rocheux fracturé. La délimitation des zones amiantées sur les photos (2D) a été le point de départ d’un troisième axe de recherche. Cette délimitation a été faite manuellement dans une première phase ; le lien entre les points 3D d’un nuage restitué par photogrammétrie et les pixels des photos utilisées pour sa restitution 3D a permis une cartographie 3D des zones amiantées connues, car identifiées in situ. La délimitation a été ensuite étendue aux zones amiantées n’ayant pas été repérées in situ par apprentissage profond (« Deep Learning »). Une méthodologie intégrant un autoencodeur (p. ex. U-Net) a été élaborée pour détecter les zones amiantifères sur les photos 2D. À nouveau, la liaison 2D-3D permise par la restitution 3D photogrammétrique a rendu possible une cartographie 3D des zones amiantées.
... The extraction of complex geological data from outcrops using photogrammetry requires the DOM, a digital 3D representation of the outcrop surface, mostly in a form of textured polygon mesh. The methodologies for extracting information from this data range from manual characterizations [62][63], to semiautomatic and automatic techniques [21,[64][65][66][67][68][69][70]. These techniques have applications in various branches of geology, where applications to geological engineering stand out with the classification and characterization of rocky massifs [71][72][73], geothreats with the determination of susceptibility to mass movements [50,74], cartography with the identification of lithological limits [75], structural geology with the use of data for analysis and structural modeling [48,[76][77][78], sedimentology [79][80], stratigraphy and reservoir characterization [81][82][83], and in geoeducation [84]. ...
Article
This paper aims to present a methodological approach for capturing information and characterizing difficult-to-access geological outcrops using unmanned aerial vehicle-based digital photogrammetric data, which has been growing in importance as a three-dimensional modeling method along with the use of 3D geomodelling, geological, stratigraphic and structural software packages, and specialized programmed algorithms in complex geological cases. In this way, it is possible to document rock outcrops, geological structures, stratification or foliation plans, geometry of outcropping lithologies, underground and surface mining works, karst systems, etc. The data obtained will then serve as a basis for the geomodelling of the geological structure of mineral deposits and oil and gas. Traditionally, the photogrammetry technique in Geosciences has been limited to simplifying and improving the work of surface mapping, topography, cartography, interferometry patterns, surface geomorphology and spectral analysis of high-resolution satellite images. However, currently, the evaluation of the discontinuities of a rock massif can be carried out, the structural domains with high precision, in a short time and in a complete way remotely, taking the information gathered in outcrops to other scenarios so that the work be interactive.
... Both calculated normal vectors are finally converted in dip-direction values. Alter- natively, linear data like slickenlines are measured by tracing a two- point line lying in the planar structure, using the same virtual compass toolbox in CloudCompare ( Thiele et al., 2017, Fig. 7a). Table A3 sum- marizes the orientation of all the structures measured in the digital outcrop model. ...
Article
Acquiring and building Digital Outcrop Models (DOM) becomes an essential approach in geosciences. This study highlights the strong potential of Structure-from-motion (SfM) photogrammetry for full-3D mapping of inaccessible outcrops, combining pictures captured from field and from unmanned aerial vehicle-embedded digital cameras. We present a workflow for (i) acquiring and reconstructing a DOM of a geometrically complex natural cave site using digital photogrammetry in a lowlight environment, (ii) georeferencing the 3D model in underground environments, (iii) identifying and characterizing the geometry of inaccessible geological structures and their tectonic kinematics (e.g., faults, joints, sedimentary bedding planes, slickenlines) for structural geology purposes. We illustrate our method by modelling a challenging case study: the main chamber of the Lorette cave (Rochefort Cave Laboratory, Belgium). First, we produced a high resolution, highly realistic model made of 395 million points cloud. This allowed to draw a detailed lithostratigraphic log of the exposed sedimentary pile, alternating decimetric carbonate mudstones with minor centimetric clay-rich layers. Secondly, we extract the orientation of brittle structures from the cave DOM which consist of joints, calcite-filled veins, fault planes with observable slickenlines and their kinematic indicators. Calcitic veins consist of tension gashes structures. Two subsets of tension gashes are distinguished based on their orientation (WNW-striking with low- vs subhorizontal dips) and morphology (planar vs en-echelon sigmoidal veins). Two faults subsets are identified: (i) a first one comprises south-dipping fault planes with mean strike-dip of N069-S42 and consist of bedding surface slip; (ii) a second one which corresponds to neoformed north-dipping faults (mean strike-dip: N279-N60). We recognize and characterize tectonic markers on fault planes directly from the high-resolution DOM (slickenlines and asymmetrical microscarps) pointing to a reverse shearing movement for all investigated faults. Based on their geometrical relations and fault-slip data, paleostress reconstruction points to a NW-SE to NNW-SSE subhorizontal compressive regime. This one is interpreted as the record of early phases of Variscan tectonics during the fold-to-fault progression. This research paper also highlights future possibilities for rapid semi-automatic interpretation of such 3D dataset for structural geology purposes as well as advances in technology and perspectives in terms of risk assessments and mitigation.
Article
This paper presents two classroom activities which have students analyze, collect, and interpret geologic data. The material is targeted at distance learning, although it can function well in the classroom as pseudo-field trips. The activity is based around Virtual Outcrop Models (VOMs), created using Structure from Motion – Multi-view Stereo photogrammetry, of two outcrops: the first is a folded outcrop in the southern Death Valley region and the second is a fractured granite in the Santa Monica Mountains. In Activity 1 students begin by analyzing the folds using 2D field photographs. They then look at the same folds using the VOM and are provided strike and dip data. Students are prompted to think about and discuss their interpretations of the shortening directions and the deformation events recorded in the outcrop. After interpreting the outcrop, they then correlate their results with regional structures. In Activity 2, students take direct fracture measurements from the VOM and interpret various fracture characteristics. Their measurements are then compared to other fracture characterizations in the literature. Through the activities students are introduced to a number of key concepts in geology, including the importance of “thinking in 3D″, data collection, interpreting stereonets, and placing results in the context of previous work.
Article
Field observations of columnar-jointed basalt lava flows at Organ Pipes National Park near Melbourne, Australia, show varying columnar cooling joint orientations. The valley-confined basalt lava flows have columnar joints forming a downward spreading fan as the joints adopt orientations perpendicular to the margins of the flow. The variation in the pattern of columnar joint orientations controls local slope failure mechanisms, including toppling and planar sliding. Manually collected field data and terrestrial laser scanning data for joint faces have been used to assess the kinematic stability of the rock slopes. The terrestrial laser scanning data obtained includes a greater number of joint face orientations than manually collected data, including data for otherwise inaccessible locations. However, the bulk data collected by TLS requires careful interpretation at sub-sites within the field of observation to enable variation in column orientation and the associated slope stability mechanisms, to be recognised. The collection of larger datasets by remote methods has the risk of smoothing out local variations. The data in this case study demonstrates that the commonly assumed simple parallel pattern of columnar joints cannot be assumed, even in large, thick lava flows such as the Organ Pipes National Park.
Article
Full-text available
The traditional study of palaeoseismic trenches, involving logging, stratigraphic and structural interpretation, can be time consuming and affected by biases and inaccuracies. To overcome these limitations, a new workflow is presented that integrates infrared hyperspectral and photogrammetric data to support field‐based palaeoseismic observations. As a case study, this method is applied on two palaeoseismic trenches excavated across a post‐glacial fault scarp in northern Finnish Lapland. The hyperspectral imagery (HSI) is geometrically and radiometrically corrected, processed using established image processing algorithms and machine learning approaches, and co‐registered to a structure‐from‐motion point cloud. HSI‐enhanced virtual outcrop models are a useful complement to palaeoseismic field studies as they not only provide an intuitive visualisation of the outcrop and a versatile data archive, but also enable an unbiased assessment of the mineralogical composition of lithologic units and a semi‐automatic delineation of contacts and deformational structures in a 3D virtual environment. L'étude traditionnelle des tranchées paléosismiques, impliquant l'enregistrement des coupes et l'interprétation stratigraphique et structurelle, peut prendre beaucoup de temps et être entachée de biais et d'inexactitudes. Pour surmonter ces limites, une nouvelle méthodologie est présentée, intégrant des données photogrammétriques et hyperspectrales infrarouges en appui aux observations paléosismiques de terrain. Comme étude de cas, cette méthode est appliquée à deux tranchées paléosismiques creusées à travers un escarpement de faille post‐glaciaire dans le nord de la Laponie finlandaise. L'imagerie hyperspectrale (HSI) est corrigée géométriquement et radiométriquement, traitée à l'aide d'algorithmes classiques de traitement d'images et d'apprentissage machine, et recalée sur un nuage de points photogrammétrique. Les modèles virtuels d'affleurements améliorés par HSI constituent un complément utile aux études paléosismiques de terrain, car ils fournissent non seulement une visualisation intuitive de l'affleurement et une archive de données facile d'emploi, mais permettent également une évaluation non biaisée de la composition minéralogique d'unités lithologiques ainsi qu'une délimitation semi‐automatique des contacts et des structures de déformation dans un environnement virtuel 3D. Die traditionelle Protokollierung und stratigraphische/strukturelle Interpretation paläoseismischer Gräben kann zeitaufwendig sein und durch Verzerrungen und Ungenauigkeiten beeinflusst werden. Um diese Einschränkungen zu überwinden, wird ein neuer Arbeitsablauf vorgestellt, der hyperspektrale und photogrammetrische Daten integriert, um feldbasierte paläoseismische Beobachtungen zu unterstützen. Als Fallstudie wird diese Methode auf zwei paläoseismischen Gräben angewendet, die über eine postglaziale Verwerfung im nördlichen finnischen Lappland angelegt wurden. Das hyperspektrale Bild wird geometrisch und radiometrisch korrigiert, mit etablierten Bildverarbeitungsalgorithmen und maschinellen Lernverfahren verarbeitet und mit einer fotogrammetrischen Punktwolke verknüpft. Hyperspektrale Aufschlussmodelle sind eine sinnvolle Ergänzung zu paläoseismischen Feldstudien, da sie nicht nur eine intuitive Visualisierung des Aufschlusses ermöglichen und ein vielseitiges Datenarchiv darstellen, sondern auch erlauben, die mineralogische Zusammensetzung lithologischer Einheiten zu ermitteln sowie Kontakte und Deformationsstrukturen in einer virtuellen 3D‐Umgebung zu analysieren. El estudio tradicional de las trincheras paleosísmicas, que implica la anotación, la interpretación estratigráfica y estructural, puede llevar mucho tiempo y verse afectado por sesgos e inexactitudes. Para superar éste hándicap, se presenta un nuevo flujo de trabajo que integra datos hiperespectrales infrarrojos y fotogramétricos para apoyar las observaciones paleoseísmicas de campo. Como caso de estudio, este método se aplica en dos trincheras paleosísmicas excavadas a través de una cornisa post‐glacial en el norte de la Laponia finlandesa. Las imágenes hiperespectrales (HSI) se corrigen geométrica y radiométricamente, se procesan utilizando algoritmos de procesamiento de imágenes establecidos y aproximaciones de aprendizaje automático, y se hacen corresponder sobre una nube de puntos derivada por fotogrametría. Los modelos de afloramiento virtual mejorados con HSI son un complemento útil para los estudios de campo paleosísmicos, ya que no solo proporcionan una visualización intuitiva del afloramiento y un archivo de datos versátil, sino que también permiten una evaluación imparcial de la composición mineralógica de las unidades litológicas y una delineación semiautomática de contactos y deformación en un entorno virtual 3D. 传统的古地震沟研究, 包括测井、地层和构造解释, 通常耗时并易受偏差和不准确性的影响。为了克服这些限制, 本研究提出新的工作流程, 整合红外高光谱和摄影测量数据, 以辅助古地震的现场观测。本研究以芬兰拉普兰北部的冰川期后断层陡坡上开挖的两块古地震沟, 作为研究案例。高光谱图像(HSI)经过几何和辐射校正, 使用既有的图像处理算法和机器学习方法分析, 并与由运动恢复结构所得点云进行配准。 HSI增强的虚拟露头模型是古地震现场研究的有效补充, 因为其不仅提供露头的直观可视化和多类型的数据存档, 而且能够对岩性单元的矿物组成, 在三维虚拟环境中进行无偏评估与半自动区分。 To overcome the limitations of the traditional study of palaeoseismic trenches, a new workflow is presented that integrates infrared hyperspectral and photogrammetric data. This method was applied on two palaeoseismic trenches excavated across a post‐glacial fault scarp in northern Finnish Lapland. The hyperspectral imagery (HSI) was corrected and co‐registered to a structure‐from‐motion point cloud. The resulting HSI‐enhanced virtual outcrop models provide an intuitive visualisation and archive of the outcrop, and enable an unbiased assessment of its mineralogical composition.
Chapter
Three major Holocene rock avalanches have sculpted the morphology of Mount Pletzachkogel (Tyrol, Austria), and rock fall processes continue to show recent activity. Coalescing sets of discontinuities and block moulds exposed in the steep and rugged limestone cliffs exemplify the broad spectrum of rotational and translational block failure modes to which the mountain is prone. As personnel safety concerns strongly limit the ability to access the 200 m high rock cliffs to make traditional structural field measurements, Unmanned Aerial Vehicle (UAV) photogrammetric surveys were performed with a compact portable multicopter. The UAV survey provided a georeferenced point cloud and Digital Terrain Model of sufficient resolution and accuracy to permit efficient extraction of structural geologic measurements by using different open source software packages. This research focuses on comparing discontinuity measurements extracted from the point cloud using manual, semi-automated, and automated techniques, to field measurements made with a geologic compass. The overall workflow of digital image processing and related structural measurement extraction is described, together with data validation procedures. The workflow described herein provides an efficient means for obtaining comprehensive and accurate data sets that mitigate personnel access constraints, are fully auditable and archivable. With increased applications of UAVs for geologic mapping and documentation, such procedures are sure to see rapidly increasing deployment, particularly in alpine terrain.
Article
Full-text available
Geological planar facets (stratification, fault, joint…) are key features to unravel the tectonic history of rock outcrop or appreciate the stability of a hazardous rock cliff. Measuring their spatial attitude (dip and strike) is generally performed by hand with a compass/clinometer, which is time consuming, requires some degree of censoring (i.e. refusing to measure some features judged unimportant at the time), is not always possible for fractures higher up on the outcrop and is somewhat hazardous. 3D virtual geological outcrop hold the potential to alleviate these issues. Efficiently segmenting massive 3D point clouds into individual planar facets, inside a convenient software environment was lacking. FACETS is a dedicated plugin within CloudCompare v2.6.2 ( http://cloudcompare.org/ ) implemented to perform planar facet extraction, calculate their dip and dip direction (i.e. azimuth of steepest decent) and report the extracted data in interactive stereograms. Two algorithms perform the segmentation: Kd-Tree and Fast Marching. Both divide the point cloud into sub-cells, then compute elementary planar objects and aggregate them progressively according to a planeity threshold into polygons. The boundaries of the polygons are adjusted around segmented points with a tension parameter, and the facet polygons can be exported as 3D polygon shapefiles towards third party GIS software or simply as ASCII comma separated files. One of the great features of FACETS is the capability to explore planar objects but also 3D points with normals with the stereogram tool. Poles can be readily displayed, queried and manually segmented interactively. The plugin blends seamlessly into CloudCompare to leverage all its other 3D point cloud manipulation features. A demonstration of the tool is presented to illustrate these different features. While designed for geological applications, FACETS could be more widely applied to any planar objects.
Article
Full-text available
The history of seafloor spreading in the ocean basins provides a detailed record of relative motions between Earth's tectonic plates since Pangea breakup. Determining how tectonic plates have moved relative to the Earth's deep interior is more challenging. Recent studies of contemporary plate motions have demonstrated links between relative plate motion (RPM) and absolute plate motion (APM), and with seismic anisotropy in the upper mantle. Here we explore the link between spreading directions and APM since the Early Cretaceous. We find a significant alignment between APM and spreading directions at mid-ocean ridges; however, the degree of alignment is influenced by geodynamic setting, and is strongest for mid-Atlantic spreading ridges between plates that are not directly influenced by time-varying slab pull. In the Pacific, significant mismatches between spreading and APM direction may relate to a major plate-mantle reorganisation. We conclude that spreading fabric can be used to improve models of APM.
Article
Digital geologic mapping is now a fully mature technology that dramatically improves field efficiency and problem solving capabilities. Basic digital mapping is just the tip of the iceberg, however, in regard to new and approaching capabilities with true 3D mapping. The key advance is the ability to easily construct high-resolution, photorealistic terrain models as a base surface for 3D mapping using Structure from Motion (SfM) photogrammetry terrain models, particularly through the aid of unmanned aerial systems (UAS). We show how these technologies can aid field visualization and discuss how developing digital field workflows and 3D visualizations will transform field studies, allowing the resolution of problems that were impossibly complex without this technology. Manuscript received 18 July 2016; Revised manuscript received 31 Jan. 2017; Manuscript accepted 12 Feb. 2017; Published online 27 Mar. 2017 © The Geological Society of America, 2017. 10.1130/GSATG313A.1
Article
Interpretation and modelling of high resolution regional geophysical data of the central Leichhardt River Fault Trough in the Mount Isa Inlier are used to determine the timing of a major basin inversion event following the development of the ca 1780-1740 Ma Leichhardt Superbasin. Inversion of the Leichhardt Superbasin formed the regional north-south trending Leichhardt Anticline during east-west shortening. The limbs of the anticline are overprinted by several east-west trending wedge-shaped, non-magnetic sub-basins filled with ca 1710 Ma Calvert and Isa superbasin successions. These relationships suggest inversion of the Leichhardt Superbasin occurred between ca 1740 and 1710 Ma. The event is also known to have affected the northern and eastern North Australian Craton. The scale of the inversion suggests it was a significant event that we have defined as the Leichhardt Event. This event requires a major tectonic driver to the east of the North Australian Craton, possibly the accretion of a micro-continental ribbon to the east of the Mount Isa Inlier. The results of this study have implication for paleogeographic reconstruction of the Australian continent during the formation of Nuna because eastern North Australian Craton faced an ocean at ca 1740-1720 Ma. The results also challenge the significance and intensity of crustal shortening associated with the ca 1600-1500 Ma Isan Orogeny throughout the western Mount Isa Inlier.
Article
Three-dimensional digital models of geological objects are relatively easy to create and geolocate on virtual globes such as Google Earth and Cesium. Emerging technologies allow the design of realistic virtual rocks with free or inexpensive software, relatively inexpensive 3D scanners and printers, and smartphone cameras linked to point-cloud computing services. There are opportunities for enhanced online courses, remote supervision of fieldwork, remote research collaboration, and citizen-science projects, and there are implications for archiving, peer-review, and inclusive access to specimens from inaccessible sites. Virtual rocks can be gradually altered to illustrate geological processes such as weathering, deformation, and metamorphic mineral growth. This paper surveys applications in a wide range of geoscience subdisciplines and includes downloadable examples. Detailed instructions are provided in the GSA Supplemental Data Repository.
Article
The eastern Coral Sea is a poorly explored area at the north-eastern corner of the Australian Tectonic Plate, where interaction between the Pacific and Australian plate boundaries, and accretion of the world's largest submarine plateau - the Ontong Java Plateau - has resulted in a complex assemblage of back-arc basins, island arcs, continental plateaus and volcanic products. This study combines new and existing magnetic anomaly profiles, seafloor fabric from swath bathymetry data, Ar-Ar dating of E-MORB basalts, palaeontological dating of carbonate sediments, and plate modelling from the eastern Coral Sea. Our results constrain commencement of the opening of the Santa Cruz Basin and South Rennell Trough to c. 48 Ma and termination at 25–28 Ma. Simultaneous opening of the Melanesian Basin/Solomon Sea further north suggests that a single > 2000 km long back-arc basin, with at least one triple junction existed landward of the Melanesian subduction zone from Eocene–Oligocene times. The cessation of spreading corresponds with a reorganization of the plate boundaries in the area and the proposed initial soft collision of the Ontong Java Plateau. The correlation between back-arc basin cessation and a widespread plate reorganization event suggests that back-arc basins may be used as markers for both local and global plate boundary changes.
Article
The routine application of digital survey technologies such as terrestrial lidar and photogrammetry to the characterization of fault and fractures in outcrop over the past decade has resulted in major advances in terms of the efficiency of discontinuity data acquisition. However, the reliance upon meshand point-cloud-based analysis approaches means that data sets obtained from these sources commonly offer heavily abstracted views of the measured fracture network due to the limited resolution of the input model. Here, we present an alternative approach that combines conventional two-dimensional (2D) image analysis with ray-tracing techniques to extract three-dimensional (3D) fracture trace maps from photogrammetrically calibrated image sequences. These 3D trace objects may be interrogated to obtain fracture network properties (trace length, intensity, and connectivity), with probabilistic methods used to estimate fracture orientation for high collinearity traces. Our approach possesses a number of advantages over existing digital surface reconstruction-based methods, with the use of a 2D pixel-based approach allowing established image-processing routines (e.g., edge detection/ connected components analysis) to be applied to the characterization of fracture and fault properties. Moreover, the innately high resolution of the input images results in practically lossless 3D fracture trace representation, limiting truncation effects. As a result, the method is capable of resolving local variability in higher-order fracture properties such as fracture intensity, which are difficult to derive using existing approaches. We demonstrate the approach on pervasively faulted Permian age exposures of the Vale of Eden Basin, UK.
Article
Brittle and ductile deformation of alternating layers of Devonian sandstone and mudstone at Cape Liptrap, Victoria, Australia, resulted in upright folds with associated fold accommodation faults and multiple fracture sets. Structures were mapped at the Fold Stack locality at Cape Liptrap using high-resolution aerial photographs acquired by a digital camera mounted on an unmanned aerial vehicle (UAV). Subsequent photogrammetric modelling resulted in georeferenced spatial datasets (point cloud, digital elevation model and orthophotograph) with sub-cm resolution and cm accuracy, which were used to extract brittle and ductile structure orientation data. An extensive dataset of bedding measurements derived from the dense point cloud was used to compute a 3D implicit structural trend model to visualise along-strike changes of Devonian (Tabberabberan) folds at the Fold Stack locality and to estimate bulk shortening strain. This model and newly collected data indicate that first generation shallowly south-southwest plunging upright folds were gently refolded about a steeply plunging/subvertical fold axis during a Devonian low-strain north-south shortening event. This also led to the local tightening of first generation folds and possibly strike-slip movement along regional scale faults. In order to distinguish fractures associated with Devonian compression from those that formed during Cretaceous extension and later inversion, we compared the five fracture sets defined at Cape Liptrap to previously mapped joints and faults within the overlying sedimentary cover rocks of the Cretaceous Strzelecki Group (Gippsland Basin), which crop out nearby. An east-southeast trending fracture set that is not evident in the Strzelecki Group can be linked to the formation of Devonian folds. Additionally, hinge line traces extracted from the Fold Stack dataset are aligned parallel to a dominant fracture set within the overlying cover sediments. This suggests that basement structures (folds and coeval parallel faults) have an important influence on fault and joint orientations within Cretaceous cover rocks.