Geologic Mapping with Scaled

3D Images
Andreas Gaich and Markus Ptsch, 3G Software & Measurement GmbH, Austria
Wulf Schubert, Institute for Rock Mechanics and Tunnelling, Graz University of Technology, Austria

Abstract
An important contribution to rock
mass characterization is related to the
mapping of discontinuities. Contactfree measuring principles tremendously
improve conventional geologic mapping
due to the ability to take measurements
fast and consistently without access and
time restrictions, as well as to provide
objective records of the rock mass.
This paper describes an approach for
rock face characterization using scaled
three-dimensional images. Two digital
photos taken with a zoom-lensed, calibrated off-the-shelf camera serve for a
three-dimensional reconstruction of the
rock face geometry. The related principles originate from computer vision,
allowing highly-flexible picture taking
and automatic image processing.
The rock face is represented on the
computer by a photorealistic spatial
representation a 3D image. From it,
measurements are taken by marking
visible rock mass features, e.g. spatial
orientations of joint surfaces and traces,
as well as areas, lengths, or positions.
Orientation measurements are depicted
in hemispherical plots together with
statistics on their spatial distribution
and rock mass parameters such as joint
spacing are provided. Two commercially
available systems have been used for
a variety of mapping projects proving
their significant support to conventional
field work.
Key words: 3D image, geologic
mapping, rock mass characterization,
photogrammetry, computer vision, discontinuity
1.

Introduction
The acquisition of geometric information on rock discontinuities is an
integral part of ground investigation
of excavation and construction works.
There is potential to enhance conventional geologic mapping due to existing
access restrictions, sampling difficulties,
human bias, and instrument errors.
However remote (contact-free) data

48 TPG NOVEMBER/DECEMBER 2008

acquisition techniques provide several

benefits, including:

efficient, detailed, and accurate

outcrop and discontinuity data
acquisition, including orientation,
geometry, and position;

enhanced worker safety (e.g. avoiding personnel at the rock face threatened by rock falls);

data archiving (permanent digital

records are acquired allowing later
assessments without time restrictions); and

data base for further analyses, such

as slope stability assessments or
numerical simulations.
The idea of using photos for geologic
mapping is straightforward and not
new. Restrictions due to limited time
and access were already identified and
addressed by several authors, such as
Linkwitz (1963), Rengers (1967), Hagan
(1980), or Crosta (1997). These previous
approaches showed the successful application of photos for supporting geologic
mapping; however efforts were too high
for reasonable and economic daily use.
Emerging digital imagery and mobile
computing power at reasonable cost
opened new possibilities as described
more recently by Roberts & Poropat
(2000), Gaich et al. (2004), or Haneberg
(2006).
In the following, a brief introduction
on the measuring principle is given, followed by the use of 3D images for geologic mapping, and ending with a brief
description of case studies.
2.

3D image generation

2.1 Background
A 3D image is a combination of a real
(digital) photograph with the geometric
information on the objects it shows. In
the actual cases the objects are rock
faces and walls. The geometry of the
exposed rock mass can be reconstructed
from digital images using methods of
Photogrammetry (Slama, 1980; Wolf &
Dewitt, 2000). Stereoscopic photogram-

metry deals with the measurement of

three-dimensional information from two
images showing the same object or surface but taken from different angles. This
principle is also referred to as Shape
from Stereo (see Figure 1). The geometric
relationship between two corresponding points in a stereoscopic image pair
is used to determine the position of
the original object point in space. This
requires accurate information on:

the image formation process of the

camera (interior orientation) see
below; and

the camera position and angular orientation when taking the pictures
(exterior orientation).

Figure 1. Shape from Stereo principle. From a pair

of images taken from different angles the geometry of an object is reconstructed. Note that the
determination of the baseline is not required
when following the computer vision approach.

Historically, the first issue led to purpose-built cameras relying on mechanically accurate imaging (so-called metric
cameras), thus being expensive. The second issue was handled either by observation of control points (points with known
co-ordinates visible in the images) or
by external measurements (Roberts &
Poropat, 2000), which is elaborate.
Beyond undertaking classical photogrammetry, computer vision in the 1990s
brought new algorithms and mathematics to the same topic (Faugeras, 1993).
The computer vision approach had sigwww.aipg.org

GEOLOGIC MAPPING WITH SCALED 3D IMAGES

nificant impacts on field handling and
image data processing as it caused
several improvements such as:

Zoom lenses can be applied. This

significantly eases the usability
and increases the flexibility in the
field;

Pictures can be taken freehand.

The relative orientation between
two images is determined fully
automatically without any control
points;

Generic 3D images can be computed fully automatically. There is

no need for explicitly knowing or
determining the interior and exterior camera orientation, i.e. generic
3D images can be generated from
pictures of virtually any digital
camera;

Metric 3D images can be generated

without information on the external orientation of the cameras. By
observing an object with known
size, a generic 3D image can be
scaled to a metric 3D image and
Simplified procedure for calibrating a camera, i.e. determination of
the interior. The interior orientation comprises the focal length, the
intersection of the optical axis with
the image plane, and a description
of the lens distortions.

2.2 Data acquisition

Utilizing the computer vision
approach data acquisition on site comprises the following steps:

Installation of reference elements:

o Either a range pole (a vertical
pole with two targets mounted
at a known distance) for local coordinate measurements or
o Installation ground control points
(surveyed in a given co-ordinate
system) for global co-ordinate
measurements

Taking two images (freehand) of the

rock face from different positions
ensuring the reference figure(s)
being visible in both images.
Measuring the length of the base
line (see Figure 1) is not required.
However, practical experience
showed that this length should be
approximately one fifth to one eight
of the mean imaging distance and
almost parallel to the strike of the
face; and

Calibration of the imaging system.

Camera calibration is crucial for

www.aipg.org

Figure 2. Several overlapping 3D images of a highway roadcut, Morrison, Colorado. All pictures were
taken freehand without surveying any camera station.

obtaining accurate results. Usually

pre-calibrated cameras are used,
thus this step can be skipped on
site.
2.3 Combination (merging) of
3D images
Practical tasks often involve dealing
with large areas, complex shapes or
high resolution, each of what go beyond
the capabilities of a single stereoscopic
image pair. In these cases, the rock wall
is to be acquired by several overlapping
3D images. They are connected to a
larger 3D model that allows for taking measurements (see Figures 2 and
3). During processing, common image
information in the overlapping regions
is used to determine a transformation
from one 3D image to the next.
3.

Assessment of 3D images for

rock mass characterization
Once a 3D image is ready, measurements can be taken directly from it. A
purpose-built 3D software component is
used that allows rotating, panning, as
well as zooming a 3D image in and out,
thus allowing a thorough inspection.
The photorealistic representation of the
rock surface together with the threedimensional shape provides a natural

impression and allows for decisions on

the geologic relevance of certain structures and their geometric properties.
In the following possible uses for geologic assessments are briefly addressed.
They are subdivided into basic mapping
features and higher level features, i.e.
rock mass related properties and evaluations.
3.1 Basic mapping features
Geometric measurements are taken
by placing graphical markers onto the
3D image. These markers denote points
or regions of interest, e.g. visible discontinuity traces or discontinuity surfaces.
All measurements taken from a 3D
image are inherently three-dimensional
in the given co-ordinate system. Pointbased, line-based, and area-based items
are available as basic mapping features.
Co-ordinates and distances
Basic magnitudes are related to surface point measurements (x,y,z coordinates) and the determination of the
Euclidean distance between arbitrarily
chosen surface points which correlates
to a virtual tape measure. By clicking
on the designated position(s) the metric
information is instantly provided.

Figure 3. Merged 3D image of Morrison site. Note that all subsequent analyses and measurements can
be performed directly in the merged 3D image.

NOVEMBER/DECEMBER 2008 TPG 49

GEOLOGIC MAPPING WITH SCALED 3D IMAGES

Individual orientations
Any location on the 3D image can be
touched with a spatial cursor that follows
the actual 3D shape of the reconstructed
surface. It changes its pointing direction
according to the actual orientation of
the surface (see Figure 4). In this way
orientation measurements are taken
which is comparable to the application
of a compass-clinometer device on a
particular location.

Figure 4. Orientations can be measured at

arbitrary locations on the 3D image. Dip
angle and dip direction are instantly provided.

Linear features
The measurement of linear rock mass
features such as joints, lithological borders, or strata is performed by marking
discontinuity traces on the 3D image. A
marked trace consists of sample points
on the surface connected to a spatial
line (a 3D poly-line). If the 3D poly-line
shows a sufficient variation in depth,
a plane is fitted automatically to the
sample points. The orientation of the
fitted plane corresponds to the spatial
orientation of the discontinuity trace
that has been marked, thus it determines
the three-dimensional orientation of the
linear feature (see Figure 5).

Figure 5. A discontinuity trace marked by a

3D poly-line. A plane is fitted through the
sampling points. Its orientation corresponds
with the orientation of the marked trace.

50 TPG NOVEMBER/DECEMBER 2008

Areas
Regions of similar geologic attributes
(e.g. lithology or same degree of fracturing) or joint surfaces are marked with
areas. An area is defined by marking a
closed poly-line on the 3D surface. The
enclosed parts of the 3D surface are used
to compute the mean orientation which is
instantly provided by dip angle and dip
direction. Figure 6 depicts an example of
a marked area and the resulting surface
normal indicating the spatial orientation
of that area.

Figure 6: Measurement of orientations at joint

surfaces. By marking points and calculating the
mean orientation of the surface normal the
orientation vector is determined. Additionally,
it delivers the size of the area.

3.2 Higher level features

From the basic measurements several
higher level features are derived with
the aim of obtaining descriptive rock
mass parameters. Basically, all rock
mass parameters based on geometric
information of rock structures can be
determined.
Structure maps
Basic features, such as joints and
areas, orientations, as well as co-ordinates, or distances are assigned to structure sets that represent geologic units,
e.g. a discontinuity set. Figure 7 shows

Figure 7: Snapshot of the software JMX Analyst

used for interactive assessment of 3D images
and the determination of descriptive rock mass
parameters.

an example of a 3D image with several

structure sets marked.
All structure sets together form a
structure map.
3.2.1
Hemispherical plots
The measurements taken from the
3D image are grouped to sets by the
operator. Each set is instantly visualized
within a hemispherical plot (stereonet)
in order to get an instant impression of
the spatial distribution of the orientation
measurements assigned to a set (see
Figure 11). Since measured structures
can be touched either within the 3D
view or from the hemispherical plot, the
proper assignment to sets is supported.
Calculated statistical parameters on
the spatial distribution of a structure set
include the spherical aperture, concentration (Fishers constant), and the cone
of confidence. The output is instantly
updated when new orientation measurements are applied.
Spacing
Structure maps inherently contain the
lengths and spacing of traces. Spacing is
referred to as set spacing, normal set
spacing, and total spacing according
to definitions given in the textbook by
Priest (1993).
The software features two possible
methods for calculating spacing. The
first one is similar to conventional scan
line mapping: the user places a virtual
scan line on the 3D image and the software calculates spacing values of the
intersected joints. The second method
is a kind of multiple scan line spacing:
traces of an entire structure set are
projected onto a reference plane. The
distances between adjacent discontinuities are determined along scan lines
perpendicular to the mean orientation.
Figure 10 shows an example of multiple
scan line spacing. In includes an automatically generated sketch that is also
used for visually reviewing the spacing
calculation.
Surface Roughness
3D images at a sufficiently high resolution can be used for obtaining discontinuity roughness values. The required
resolution for measuring discontinuity roughness is not a fixed value, but
depends on the scale of analysis (looking at waviness requires less resolution
than analyzing roughness). Figure 8,
for instance, shows a roughness profile along a discontinuity plane. The
point density of the corresponding 3D
image is approximately1.2 pts/cm2. The

www.aipg.org

GEOLOGIC MAPPING WITH SCALED 3D IMAGES

obtained roughness profiles can then
be compared with standardized profiles
(ISRM, 1978).

Figure 8: Roughness profile evaluated from a

3D image. The point density is approximately
1.2 pts/cm2

4.

Applications
In the following, case studies using the commercial systems
ShapeMetriX3D
and
JointMetriX3D are briefly touched
upon in order to highlight the
capabilities of the 3D imaging technology.

In conventional tunnelling, face mapping has to be performed quickly. Two

photos of the tunnel face can be taken
within a minute without significant
disruption of excavation works (cf. Gaich
et al. 2004). This provides the geologists
with more time on site for the analysis of
other parameters as the actual geometry
is already captured.
Assessments (see Figure 12) are performed on the computer without further
time and access restrictions. Subsequent
3D images of tunnel faces (see Figure
13) represent an objective, reproducible record of the rock mass conditions

Figure 12: 3D image of a tunnel face (cross

section about 25 m2) with main features
mapped.

ing system together with a reference figure allows generating a

scaled 3D image of the observed
rock face for both visualizing the
actual conditions and obtaining
accurate measurements on rock
structures. 3D imaging based on
computer vision is easy to use, can
4.1 Rock slope
be handled by one person, involves
The stability of a rock slope
only light-weight equipment, and
with a height of about 150 m was
avoids operator hazards due to its
assessed. Several parts were inacremote application.
Figure 10: Computer generated sketch of joint traces for
cessible, so contact-free measureone
set
together
with
statistics
for
determining
normal
It can be used from close range
ments proved to be a proper way
spacing and joint frequency
(below 1 m) to large distances
to gather reasonable quantitative
(beyond 1,500 m) on the surface
geometric information on the disand underground. Applications
continuity network and the free
include documentation of the encounsurface.
tered rock mass conditions, support of
A highly detailed 3D image (70 megageologic mapping and rock mass characpixels) was generated in order to allow
terization, or discontinuity and keyblock
a geologic assessment also of smaller
analysis. Practitioners and engineers
structures. Figure 9, Figure 10 and
may explore additional applications of
Figure 11 show the achieved results of
this technology.
the 3D imaging and assessment activi3D imaging with according assessties using the ShapeMetriX3D software.
ment software will have a significant
Within the software measured orientaimpact on the current analysis and
tions are instantly displayed together
design practice in rock engineering.
with statistics on their spatial distriHowever, these tools are intended as a
bution. Furthermore spacing and joint
support and not a substitute to convenlength statistics are provided.
tional field work.
4.2 Tunnel face mapping

Figure 11: Lower hemisphere equal-area projection polar plot of identified discontinuity
sets.

encountered during excavation delivering a good data base for any later review
of a project.
5.
Figure 9. Application of 3D imaging technology for the analysis of a 150 m rock slope.

www.aipg.org

Conclusion
3D imaging is a powerful technology
for the documentation and characterization of exposed rock faces. Using
only two images of a calibrated imag-

Figure 13: Subsequent 3D images of a drift

tunnel excavation in an underground marble
mine.

NOVEMBER/DECEMBER 2008 TPG 51

GEOLOGIC MAPPING WITH SCALED 3D IMAGES

Ref er en ces
Cr ost a G. (1997) Evaluat ing Rock M ass
Geomet r y Fr om Phot ogr ammet r ic
I mages. Rock M echanics and Rock
Engineer ing, 30(1): 35-38.

I SRM (1978). Suggest ed met hods for t he

quant it at ive descr ipt ion of discont inuit ies in r ock masses. I nt . J. Rock
M ech. M in. Sci. & Geomech. Abst r .
15: 319-368.

Slama, Ch. C. (ed.) (1980) M anual

of Phot ogr ammet r y. 4t h edi t i on.
Am er i can
Soci et y
of
Phot ogr ammet r y, Fal l s Chur ch,
VA.

Fauger as, O. (1993) Thr ee-Dimensional

Comput er Vi si on. M I T Pr ess,
Bost on, M A.

L i nk wi t z K . (1963) Ter r est r i schphot ogr ammet r ische K luft messung.

Rock M echanics and Engineer ing
Geology I : 152-159.

Wolf, P.R. & Dewit t , B.A. 2000. Element s

of Phot ogr ammet r y. Thir d Edit ion.
M cGr aw-H ill, Bost on.

Gai ch, A., Schuber t , W. & Pt sch,

M . (2004) Repr oduci bl e r ock
mass descr ipt ion in 3D using t he
Joint M et r iX3D syst em, Pr oc. of t he
I SRM Regional Symposium Eur ock
2004 & 53r d Geomechanics Colloquy,
Salzbur g, Aust r ia, pp. 61-64.
Gaich, A., Pt sch, M . & Schuber t , W.
(2006) Acquisit ion and assessment
of geomet r ic r ock mass feat ur es by
t r ue 3D images. I n ARM A Golden
Rocks 2006 50 Year s of Rock
M echanics, Golden, Color ado, 17-21
June 2006, Paper 06-1051.
H agan T.O. (1980) A Case for Ter r est r ial
Phot ogr ammet r y i n Deep-M i ne
Rock St r uct ur e St udies. I nt . J. Rock
M ech. M in. Sci. 17: 191-198.
H aneber g, W.C. (2006) 3-D Rock M ass
Char act er izat ion Using Ter r est r ial
Di gi t al Phot ogr ammet r y. AEG
News 49(4): pp. 12-15.

52 TPG NOVEMBER/ DECEMBER 2008

Pt sch, M ., Schuber t , W. & Gaich, A.

(2006) K inemat ical analyses of r ock
blocks suppor t ed by 3D imaging.
I n ARM A Golden Rocks 2006 50
Year s of Rock M echanics, Golden,
Color ado, 17-21 June 2006, Paper
06-1079.
Pr iest , S.D (1993) Discont inuit y Analysis
for Rock Engineer ing. Chapman and
H all, L ondon.

R ev i ew ed b y
A I P G Associ at e
E d i t or s: N eil l Ridgley, CPG-05138,
Scot t Tiller , CPG-10016, and Ed Balt zer ,
CPG-08861.
Dr . Andr eas Gaich holds a M ast er s
and PhD degr ee fr om Gr az Univer sit y
of Technology, Aust r ia. Since 2003 he is
t he M anaging Dir ect or of 3G Soft war e &
M easur ement , a company specialized in
3D imaging syst ems for geological and
geot echnical applicat ions.

Renger s
N.
(1967)
Ter r est r i al
Phot ogr ammet r y: A Valuable Tool
for Engineer ing Geological Pur poses.
Rock M echanics and Engineer ing
Geology V: 150-154.
Rober t s, G. & Por opat , G. (2000)
H ighwall joint mapping in 3D at
t he M our a mine using SI ROJOI NT.
Bowen Basin Symposium 2000 Coal
and M ining The New M illennium,
Rock hampt on.

www.aipg.org