Content uploaded by Thomas Breuer
Author content
All content in this area was uploaded by Thomas Breuer on Feb 15, 2018
Content may be subject to copyright.
NEW APPROACHES
Validation of Two Independent Photogrammetric Techniques for Determining
Body Measurements of Gorillas
JORDI GALBANY
1
*, TARA S. STOINSKI
2,3
, DIDIER ABAVANDIMWE
2
, THOMAS BREUER
4
,
WILLIAM RUTKOWSKI
5
, NICHOLAS V. BATISTA
6
, FELIX NDAGIJIMANA
2
,AND SHANNON C. MCFARLIN
1,7
1
Department of Anthropology, Center for the Advanced Study of Human Paleobiology, The George Washington University,
Washington DC
2
Dian Fossey Gorilla Fund International, Atlanta, Georgia
3
Zoo Atlanta, Atlanta, Georgia
4
Global Conservation Program, Wildlife Conservation Society, Bronx, New York
5
Department of Physics, The George Washington University, Washington DC
6
Department of Mechanical and Aerospace Engineering, The George Washington University, Washington DC
7
Division of Mammals, National Museum of Natural History, Smithsonian Institution, Washington DC
The ability to accurately measure morphological characteristics of wild primates in the field is
challenging, yet critical for understanding fundamental aspects of their biology and behavior. Recent
studies have shown that digital photogrammetry can be used to non-invasively measure morphological
traits of wild primates, as it allows for the determination of geometric properties of objects remotely
from photographic images. We report here on a rare opportunity to test this methodology by comparing
measurements obtained directly from living great apes to those obtained from photographs. We test the
accuracy and precision of two independent photogrammetric techniques, employing the use of parallel
lasers and a distance meter, respectively, for obtaining measurements of static objects and captive
western lowland gorillas (Gorilla gorilla gorilla)(n¼4) at Zoo Atlanta. For static objects, the mean
percent error between corresponding measurements collected by the same observer directly versus
using photogrammetry was 0.49–0.74% for the parallel laser method and 0.62–0.76% for the distance
meter method. For gorillas, mean percent error between corresponding direct and remote measure-
ments was 2.72–5.20% for the parallel laser method and 2.20–7.51% for the distance meter method.
Correlations between direct measurements and corresponding parallel laser and distance meter
measurements of gorillas were highly significant with R
2
values and slopes approaching 1.0 (parallel
lasers: R
2
¼0.9989, P<0.0001; distance-meter: R
2
¼0.9990, P<0.0001). Further, variation between
measurements of the same targets collected repeatedly by the same observer, and between different
observers, was uniformly low across methods (CV, range ¼0.003–0.013). While errors are slightly
higher for the distance meter technique, both methods show great promise for addressing a wide range
of questions requiring the non-invasive collection of morphological data from wild primates. Am. J.
Primatol. © 2015 Wiley Periodicals, Inc.
Key words: photogrammetry; parallel laser; distance meter; gorilla
INTRODUCTION
The ability to accurately measure morphological
traits of wild primates in the field is critical for
understanding fundamental aspects of their biology,
including developmental aspects of life history,
sexual dimorphism, correlates of reproductive suc-
cess, and health. However, as this can be logistically
challenging and invasive, most available data comes
either from captive primates or measurements of
wild primates after death. Several studies have
developed non-invasive approaches, such as the use
of a weight scale, to allow for collection of body mass
data from wild animals [e.g., Altmann & Alberts,
2005; Johnson, 2003; Pusey et al., 2005]. However,
Contract grant sponsor: The Leakey Foundation;
contract grant sponsor: The Wenner Gren Foundation;
contract grant sponsor: Center & Institute Facilitating Fund
and Signature Program funding to the Center for the Advanced
Study of Human Paleobiology, The George Washington Uni-
versity; contract grant sponsor: Columbian College of Arts and
Sciences; contract grant sponsor: National Science Foundation;
contract grant number: BCS 1520221.
Correspondence to: Dr. Jordi Galbany, Department of Anthro-
pology, Center for the Advanced Study of Human Paleobiology,
The George Washington University, 800 22nd Street NW, Ste
6000, Washington DC 20052. E-mail: jgalbany@gwu.edu
Received 23 August 2015; revised 29 October 2015; revision
accepted 24 November 2015
DOI: 10.1002/ajp.22511
Published online XX Month Year in Wiley Online Library
(wileyonlinelibrary.com).
American Journal of Primatology
© 2015 Wiley Periodicals, Inc.
this is not possible in all settings, as it may interfere
with the natural behavior of study animals.
Photogrammetry is a technology that allows for
the determination of spatial measurements of objects
indirectly, or remotely (i.e., without physical con-
tact), from photographic images [Mikhail et al., 2001;
Walker & Alspaugh, 2013]. Advancements in this
field have been driven by the need to obtain reliable
geospatial information from the earth’s surface
[McGlone, 2013], including more recent applications
that entail geospatial mapping of archaeological or
paleontological sites [Bates et al., 2008; Breithaupt &
Matthews, 2001]. Photogrammetry provides a means
of generating a permanent photographic record of
morphology, whether of inanimate objects or biologi-
cal specimens [Falkingham, 2012; Mallison & Wings,
2014], and for collecting measurements of study
animals in the field [e.g., Bergeron, 2007; Brager and
Chong, 1999; Deakos, 2010; Durban & Parsons, 2006;
Ireland et al., 2006; Jaquet, 2006; Morgan & Lee,
2003; Shrader et al., 2006]. Photogrammetry has
also been applied to humans [Gavan et al., 1952;
Geoghegan, 1953; Tanner & Weiner, 1949] and to
wild primates, the latter to measure features such as
sexual swelling size [Deschner et al., 2004; Domb &
Pagel, 2001; Emery & Whiten, 2003; Fitzpatrick
et al., 2014], sagittal crest size and other cranial
features [Breuer et al., 2007, 2012; Caillaud et al.,
2008], tail length [Rothman et al., 2008], and body
segment lengths [Barrickman et al., 2015; Bergh€
anel
et al., 2015; Brazeau et al., 2013; Breuer et al., 2007,
2009, 2012; Kurita et al., 2012; Lu et al., 2013].
Previous efforts to validate photogrammetric
techniques applied to studies of primate morphology
often incorporate comparisons of measurements
obtained directly from static test objects against
those obtained remotely from photographs [e.g.,
Barrickman et al., 2015; Breuer et al., 2007; Roth-
man et al., 2008]. However, there have been few
opportunities to directly test the accuracy of photo-
grammetry for collecting linear measurements of
living primate subjects. Rothman et al. [2008] tested
the accuracy of a parallel laser method, in which
paired lasers separated by a known distance are
projected into the photographic imaging plane as a
scale, for measuring tail length of red colobus
monkeys in Kibale National Park, Uganda. The
mean percent error between direct and photogram-
metric measurements of tail length reported in the
latter study was 1.7%, with a maximum error of 5.0%
for any single measurement. Further, they found
that direct and estimated tail lengths were highly
correlated [See also Barrickman et al., 2015 for an
application to howler monkeys].
These results suggest that photogrammetry is a
promising method for obtaining body measurements
from arboreal primates in field settings. However,
opportunities to validate photogrammetric techni-
ques using living great ape subjects, which are
larger-bodied, often observed from the ground, and
differ in their physical characteristics (e.g., length of
body hair), are rare due to the invasive nature of such
tests [but see Machanda et al., 2015].
In the current study, we test accuracy and
precision of two independent photogrammetric
methods, incorporating the use of a distance meter
[following Breuer et al., 2007] and parallel-laser
apparatus [following Rothman et al., 2008], imple-
mented under the same conditions for collecting
measurements of both static/inanimate objects and
western lowland gorillas (Gorilla gorilla gorilla)
housed at Zoo Atlanta. Since the Zoo Atlanta gorillas
have been trained to present themselves for body
measurements during veterinary examinations, this
provides a rare opportunity to compare the quality of
photogrammetric measurements to those collected
directly from living great apes. Furthermore, if both
techniques can be shown to produce accurate and
precise estimates of body size in gorillas, this
expands potential for future comparative studies
across populations where logistical constraints (and
thus the applicability of one method over another)
may vary considerably.
DESCRIPTION
We tested the application of two independent
methods for measuring the linear dimensions of static
objects and living gorillas from photographs. These
methods entail the use of a commercially available
digital single-lens reflex (DSLR) camera system, and
incorporation of a distance meter [following Breuer
et al., 2007] or parallel laser apparatus [following
Rothman et al., 2008], respectively. The general
principles of these methods are reviewed below,
followed by a discussion of sources of error.
Method 1: Distance Meter
We tested the distance meter method as de-
scribed by Breuer et al. [2007] and applied to wild
western gorillas observed at Mbeli Bai, Republic of
Congo [also see Breuer et al., 2009, 2012; Caillaud
et al., 2008]. This method relies on collecting accurate
measurements of the distance between the target
object plane, a plane that is oriented perpendicular to
the optical axis and contains all target landmarks for
measurement, and the lens (i.e., object distance).
According to this method, the size of an object (o)is
determined by the ratio of the principal distance of
the lens and the object distance:
D=f¼o=p;ð1Þ
where fis the principal distance of the lens (termed
focal length when the lens is focused at infinity); Dis
the object distance, measured from the camera lens
Am. J. Primatol.
2/Galbany et al.
to the target object; and pis the pixel size of the object
in the photo. prepresents the number of pixels
comprising the linear distance of the object in the
photo, multiplied by pixel length—an intrinsic
parameter of the camera sensor itself.
In a single, thin lens system, the object distance
can be modeled as the distance from the target object
plane to a plane that is perpendicular to the optical
axis and passes through the center of the lens. For a
lens focused on an object at infinity, the principal
distance (or focal length) is the distance from the
center of the lens to the point at which light rays
converge to form a focused image (i.e., the focal
plane), calculated according to the lens maker’s
equation from the radius of curvature of both lens
surfaces and the index of refraction [Fiete, 2013;
Mikhail et al., 2001]. However, in reality, lenses
typically used with commercially available DSLR
camera systems have multiple elements inserted
along the optical axis and cannot be appropriately
modeled as a thin lens system [Bentley & Olson,
2012; Kingslake, 1992]. Thus, the lens system is
redefined such that principal distance and object
distance are measured in reference to two principal
planes, each oriented perpendicular to the optical
axis. Assuming the same index of refraction on both
sides of the lens, the principal planes intersect the
optical axis at nodal points; in the absence of lens
distortion, a principal ray passing through the front
(incident) node will emerge from the rear (emergent)
node without changing its angle with respect to the
optical axis, though it will be displaced. In this
system, the object distance is defined in reference to
the front principal plane (or incident node); when
focused on an object at infinity, the principal distance
is defined as that distance between the rear principal
plane (or emergent node) and the focal plane
[Greivenkamp, 2004; Kerr, 2004].
Method 2: Parallel Lasers
A second method relies on the projection of dual
parallel lasers to calibrate photographic images, and
has been applied to studies of red colobus monkeys
[Rothman et al., 2008], howler monkeys [Barrickman
et al., 2015], and to other mammals [e.g., Bergeron,
2007; Durban & Parsons, 2006]. Photographs are
collected using a DSLR camera mounted to an
aluminum base, configured with a housing structure
that projects parallel lasers into the target object
plane. This method is based on the principle that when
laser beams are aligned parallel to one another, they
project laser points onto a target that remain
equidistant irrespective of changes in distance from
their origin [Rothman et al., 2008]. If measurement
landmarks on the target object are positioned in a
plane (i.e., the object plane, as defined above) that
is perpendicular to the projection plane of the
laser beams, and the distance separating each of the
lasers is known, this provides a scale that allows for
calibrationof photographs for measurement purposes.
Sources of Error
The quality of data obtained using photogram-
metric methods is dependent upon different sources of
error, which determineaccuracy and precision [Butler
et al., 1998; Cooper & Cross, 1988]. Accuracy can be
defined as the degree of agreement between remote
measurements and accepted reference values, which
in this case are corresponding measurements ob-
tained directly from the target object. Accuracy is
dependent upon systematic errors generated as from
the photogrammetric apparatus, image collection
and/or measurement procedures. Precision is a
function of random errors in the imaging and/or
measurement procedure, and can be defined as the
degree of agreement among repeated observations of
the same object made under identical conditions.
Systematic error may be introduced as a result of
the properties of the apparatus. The properties of the
digital sensor itself, including pixel size, density, and
sensor size, can all influence image quality, particu-
larly under low light conditions, and the ability to
resolve closely spaced landmarks on the target object
[Butler et al., 1998; McGlone, 2013; for a helpful
online resource, see https://photographylife.com/the-
benefits-of-a-high-resolution-sensor]. Lens aberra-
tions, or departures from the theoretical predictions
of geometrical optics, can also significantly impact
image quality [Bentley & Olson, 2012; Kingslake,
1992]. Though advances in photographic lens design
may reduce aberrations, lens distortion remains a
concern, as this may introduce error in measure-
ments obtained from photographs [also see Fiete,
2013; Mikhail et al., 2001]. Lens distortion results
from the curved geometry of lenses, such that points
in the optical field are projected as being closer or
farther from the center than their true distance. This
can be detected as a deviation from rectilinear
projection, whereby straight lines in the field of
view are projected as straight lines in the photo-
graphic image. Radial distortion increases from the
center of the optical field, and is greatest for objects
located at the periphery. Available software pack-
ages (e.g., Adobe Photoshop CS6 Lens Correction
filter) allow for post-acquisition algorithmic trans-
formations of digital images to correct for lens
distortions, though improvements in accuracy should
be evaluated [Hugemann, 2010]. However, the
degree of distortion can vary depending on lens
construction, and should be a factor in choice of
lenses for use in photogrammetry.
An additional source of systematic error arises
from the problem of “focus breathing,”which
specifically impacts methods that rely on principal
distance as a parameter in measurement calcula-
tions (i.e., the distance meter method above). In
Am. J. Primatol.
Photogrammetry and Body Measurements in Gorillas /3
compound lenses with multiple optical elements, the
positions of these optical elements “float”as the lens
is focused, and this movement changes the principal
distance of the lens in relation to the focal distance.
Although this effect tends to be greater in zoom
lenses compared to prime lenses, the latter are not
free of this effect. The manufacturer specified “focal
length”for prime lenses is the nominal principal
distance for an object focused at infinity. However,
principal distance is reduced as object distance is
reduced [Tubbs & Ito, 2002].
Both the distance meter and parallel laser
method are subject to systematic error due to
improper alignment [e.g., Bergeron, 2007; Rothman
et al., 2008]. If projection of the distance meter onto
the target object is not oriented parallel to the
ground (for terrestrial animals), distance will be
overestimated (see Equation 1). Also, to ensure the
accuracy of measurements derived using the laser
method, it is critical that paired lasers are oriented
parallel to one another. Otherwise, paired laser
distance changes with increasing object distance,
hence impacting calibration of acquired images for
measurement. Finally, both methods require that
the target object plane (i.e., defined by the land-
marks selected for measurement) is positioned
perpendicular to the optical axis of the camera. If
the target object plane is rotated toward or away
from the optical axis, this introduces cosine errors;
the object distance derived from photogrammetry
will underestimate the true distance, and this error
will increase in magnitude with increasing rotation
of the target [Bergeron, 2007]. In practice, however,
rotations up to 10 degrees can be accommodated
with a reduction in target object length of 1.5%. As
target object rotation surpasses 25 degrees, error
exceeds 10% [1–cos a;alsoseeBarrickmanetal.,
2015; Jaquet, 2006].
Repeated measurements of the same target
object can also generate different values as a
result of random errors during the imaging and/or
measurement procedure. These random errors
may be observed when a single individual conducts
repeated observations of the same target (i.e.,
intraobserver error), or in the variation observed
among measurements of the same target object
collected by multiple individuals (i.e., inter-
observer error). These random errors may result
from variations in the alignment of the optical axis
relative to the target object plane during imaging.
Moreover, when the target (i.e., gorilla) is moving,
the obtained distance may be also slightly different
than the real object-distance. Finally, both random
and systematic errors may also occur during
subsequent digital photograph processing and
analysis. One’s ability to clearly identify landmarks
in the photographs, necessary to minimize intra-
and inter-observer errors, may be impacted
by properties of the camera sensor, including
resolution, illumination of the subject, image
compression generated by measurement software,
and characteristics of the animal itself [Butler
et al., 1998; McGlone, 2013].
EXAMPLE
Ethical Statement
Our protocols comply with the American Society
of Primatologists Principles for the Ethical Treat-
ment of Non-human Primates, and received exemp-
tion by the Institutional Animal Care and Use
Committee of The George Washington University
and Zoo Atlanta’s Scientific Advisory Committee.
Gorilla Study Subjects
We collected data over 14 days in August–
September of 2013 at Zoo Atlanta in Atlanta,
Georgia. Adult gorillas are trained to present
themselves for examination by Zoo Atlanta staff.
During these examinations, gorillas are separated
from zoo staff by a vertical steel mesh partition;
gorillas are trained to present designated parts of
their body flat against the mesh, while staff on the
opposite side of the partition collect measurements of
these body parts as projected across the mesh, using
a tape measure.
The current study incorporates data from three
adult females (Lulu, Sukari, and Kuchi) and one
silverback male (Charlie) from two different social
groups, for whom both direct measurements and
photogrammetric measurements could be obtained.
As there is variation among individuals in their
training for veterinary examinations, these individu-
als were also chosen as they performed most reliably
during measurement procedures. Our four study
subjects were housed in two separate enclosures
characterized by naturalized environments that pro-
mote behaviors similar to those observed in the wild.
Each provided opportunities for an observer standing
outside the enclosure to view and collect photographs
of the gorillas from a distance of 7–13 m. We also
collected direct and photogrammetric measurements
of static objects, as described further below.
Gorilla Body Measurements
For photogrammetric measurements, individu-
als were photographed in lateral view (norma
lateralis) while in their exhibit enclosures, from
an observer distance (parameter Din Equation 1) of
7–13 m. We examined three dimensions that could
be reliably obtained during direct measurements
(given the logistical constraints of obtaining meas-
urements across a vertical mesh partition) and
for which landmarks could be readily identified
from photographs. Two of these measurements
were obtained from the head in norma lateralis
Am. J. Primatol.
4/Galbany et al.
orientation: (1) distance from the skin overlying the
most anterior margin of the external auditory meatus
to the skin covering the anterior-most projection of
the ipsilateral supraorbital torus (EAR-TORUS);
and (2) distance from the skin overlying the most
anterior margin of the external auditory meatus to
the superior-most projection of the sagittal crest
(EAR-CREST). The third measurement obtained was
the arm segment length (ARM), measured in lateral
view, as the distance from the top rounded contour of
the shoulder to the most distal protuberance of the
elbow (Fig. 1). We selected only those photos that
showed subjects in norma lateralis orientation, with
no apparent rotation of the subject in the coronal
plane (i.e., towards or away from the camera).
The Apparatus
Digital photographs were collected using a full-
format Nikon D800 digital SLR camera (36.3 MP
resolution) with a Nikon AF Micro-Nikkor 200mm
f/4 D IF-ED lens, a system with excellent marks in
tests of low light performance, optical distortion
and image sharpness (http://www.kenrockwell.com/
nikon/200mm-micro.htm). A Leica Geosystems DISTO
E7400Laser Distance Meter (reported accuracy:
1 mm at a range up to 80 m) was used to collect
measurements of object distance.
The design of our parallel laser apparatus follows
Rothman et al. [2008] and Bergeron [2007], with a few
modifications (see Fig. 2). Like Rothman et al. [2008],
we used AGLM2 green laser modules, which in our
experience showed greatest contrast against the
subject and were more easily visualized during
daylight. The AGLM2 modules (wavelength 532 nm,
maximum output power <5 mW) were produced by
Apinex (Montreal, Canada), and are rated as Class
IIIa according to the U.S. Food and Drug Admin-
istration’s Center for Devices and Radiological Health.
We modified previously published designs for
the laser apparatus in the following ways. First, the
lasers were mounted in a vertical stainless steel plate
rather than aluminum, allowing for thinner material
to be used while still maintaining the structure and
strength of the apparatus. We also added a housing
behind the front plate to enclose and protect the laser
hardware and wiring from exposure to the outside
elements. The stainless steel housing was designed
with an array of openings to allow for increased
convective cooling of hardware and the overall
housing. Each laser was connected to a battery
pack that held two AA batteries and an on/off switch.
The battery packs were mounted directly to the
stainless steel housing to form a compact design.
As we found small-scale variation among lasers
in the projection angle of the beam as it exits the
module, each laser module was fixed into the vertical
plate using aluminum collars that allowed minor
adjustments in the orientation of each module using
one set screw and three hex screws. Each hex screw
was machined such that it formed a paired concave-
convex joint with its aluminum collar. The base of
each aluminum collar, and its corresponding surface
on the vertical plate, was also machined to a paired
concave–convex contour to allow for more range of
motion in aligning the lasers to parallel orientation
(see “Alignment of parallel lasers,”below) (Fig. 2).
Second, our laser housing design allows for the
mounting of three, rather than two, lasers—each
separated by a distance of 4.0 cm and positioned at
three corners of a square. The paired laser distance
Fig. 1. Measurements considered: EAR-TORUS (A), EAR-CREST
(B) and ARM (C). The gorilla IDs are Charlie (silverback male, top)
and Kuchi (adult female, bottom).
Am. J. Primatol.
Photogrammetry and Body Measurements in Gorillas /5
represents a compromise. Extrapolation beyond this
distance is required when measuring large objects,
which can magnify any small errors in the calibra-
tion. However, if paired lasers are separated by too
great a distance, this precludes our ability to
measure small features (e.g., on the head of infants).
We designed our apparatus to accommodate
parallel laser alignment in both x(horizontal) and
y(vertical) axes to address the problem of parallax
[Bergeron, 2007; Rothman et al., 2008]. As Rothman
et al. [2008] noted, when two parallel lasers are
positioned vertically with respect to one another and
the apparatus is rotated upwards to photograph an
arboreal target, difference in path length between
the top and bottom laser beams alters the distance
between laser points projected on the target itself
[see Fig. 2 in Rothman et al., 2008]. Rothman et al.
[2008] described a method to correct for parallax in
such cases, where the angle formed between the
projected laser beams and the horizontal is deter-
mined with use of a clinometer. However, if parallel
lasers are aligned in a horizontal plane with respect
to one another in such a scenario, there should be no
difference in their path lengths and a correction
factor is not needed, assuming there is no tilt toward
or away from the target object plane. In our case,
testing both parallel laser and distance meter
methods in the field, it was logistically challenging
to collect clinometer readings consistently in all
cases. Thus, in an attempt to streamline the
technique, designing the apparatus with three lasers
allowed us to calibrate each photograph in both
vertical and horizontal axes, as necessary, depending
upon the orientation of the target object.
Procedure
Alignment of parallel lasers
Prior to each photography session, the position of
the lasers was finely adjusted until vertical and
horizontal pairs were each separated by a distance of
4.0 cm. A known size square-shape target (all four
cornersseparated bya distance of 4.0 cm in vertical and
horizontal planes) was positioned on a wall, and the
lasers were projected onto the target from a distance of
13 m, parallel to each other and to the ground to avoid
parallax.Eachlaserbeamwasfinelyadjustedbyhandto
each corner of the target and then secured in place. To
determinetheerrorinthealignmentinbothxandyaxes,
a photo of the target was then obtained with the three
parallel lasers projected into the photographic image
plane to serve as a scale.
Calculating object distance (D) and principal
distance (f) for the distance meter method
Though we used a prime lens, our measurements
may still be impacted by the phenomenon of focus
breathing described above; as lens elements float to
accommodate focusing on objects at different distance,
both the position of principal planes and the principal
distance also change. Here, it was necessary to define
object distance using a fixed landmark. Thus, object
distance in our protocol was defined as that distance
from the target object plane to the front vertex of the
lens. To determine principal distance at different
object distances, we obtained seven photographs of a
static object of known dimensions at different dis-
tances (6–16 m), and solved for principal distance
using Equation 1, above. We then calculated a linear
regression between the object distance and the
estimated principal distance (f¼0.0002 Dþ204.3;
R
2
¼0.891). This equation was used to calculate the
principal distance for all photographs measured here.
Criteria for selection of photographs
To minimize error described above, the following
criteria determined selection of appropriate photo-
graphs for measurement. (1) Landmarks were in
focus and could be easily identified; (2) Projected
laser points were located on the body, in the same
plane as the measurement landmarks; (3) The target
object plane appeared from visual inspection to be
perpendicular to the optical axis of the lens and
projection path of the lasers, to minimize parallax
error; (4) The gorilla was located in the center of the
image (i.e., not at the periphery, where lens distor-
tion effects are greatest) and photographed while
sitting or standing still, as movement of the gorilla
toward/away from the lens introduces error in object
distance determination.
Collection of measurements from photographs
Photographs of static objects were collected by
two observers (JG, DA) according to the procedures
Fig. 2. Photogrammetric kit, including a digital SLR camera and
housing for three parallel lasers mounted onto a steel base.
Separate AA battery packs for each laser module are mounted on
the outside of the laser housing (A). The parallel laser housing
allows for positioning of four laser modules for flexibility. Only
three modules are used here. (B). Close-up view of the aluminum
collar surrounding each laser module (C).
Am. J. Primatol.
6/Galbany et al.
described above; only the first author had the
opportunity to collect photographs of Zoo Atlanta
gorillas. Pixel length measurements of targets were
collected from digital photos by the first author for
most calculations; a second observer (DA) collected
repeated measurements from a subset of photos for
determination of inter-observer error. All measure-
ments were conducted using the “measure tool”in
ImageJ 1.47v [Abramoff et al., 2004]; measurements
were collected with single pixel resolution, using a
mouse to identify target landmarks representing the
start and end points of the measure line in ImageJ.
For the laser method, the distances between hori-
zontal and vertical paired lasers were measured; the
average of these distances was used to set the scale in
each photograph.
Testing Precision and Accuracy of
Photogrammetric Methods
We tested the precision of our measurements,
which is impacted by random errors in the imaging
and/or measurement procedure. Intra-observer error
was determined in two ways. Within-photo error was
determined by collecting three repeated measure-
ments from a subset of the photos (e.g., one photo for
each gorilla) on three separate occasions. Further,
each individual/target object was photographed four
times per measurement for determination of between-
photo measurement error. (All photographs and
measurements of gorillas for calculations of intra-
observer error were performed by the first author,
JG). We also determined inter-observer error in two
ways: (1) a second observer (DA) re-measured a subset
of those gorilla and object photos described above to
estimate within-photo error among observers; and (2)
photographs of static objects were collected by two
different observers (JG and DA) and measured by JG
to estimate error associated with different photogra-
phers. We report coefficients of variation (CV), as
previously used in Breuer et al. [2007], and performed
Wilcoxon Tests to determine differences between
measurements by the two observers.
Finally, to validate the method, we report here on
the accuracy of photogrammetric measurements
compared to measurements collected directly from
the same structures—for both static objects and
western gorillas observed at Zoo Atlanta. Static object
measurements were collected with a measuring tape
having a resolution of 0.1 cm. All direct measurements
of gorillas were obtained by Zoo Atlanta staff using a
flexible tape measure with 0.5 cm resolution, based on
the Zoo’s established measurement protocols and
what has been deemed feasible during past veterinary
examinations. To examine the relationship between
direct and photogrammetric measurements, we cal-
culated the percent of error between corresponding
sets of measurements, and performed ordinary least
squares linear regression analysis. Finally, we used
regression analysis to determine whether error
between direct and photogrammetric measurements
is significantly biased by object distance.
Results: Measurements of Gorillas
The parallel lasers were aligned five different
times during the period of photo acquisition from Zoo
Atlanta gorillas; the average error between laser and
direct measurements of the calibration target was
1.01% for the horizontal axis (range ¼0.64–1.30%),
and 1.04% for the vertical axis (range ¼0.28–1.91%).
Repeated direct measurements of the gorillas
collected by a single observer (Zoo Atlanta Primate
Staff –RPG) yielded a mean CV ¼0.02 (range ¼0.01–
0.03) for ARM, mean CV ¼0.03 (range ¼0.02–0.05)
for EAR-CREST, and mean CV ¼0.05 (range ¼0.02–
0.09) for EAR-TORUS (Table I).
Precision
To estimate between-photo error, target lengths
were obtained by a single observer using photogram-
metric methods from four different photos collected
per individual gorilla. CVs were similar for the
parallel-laser and distance meter methods, and also
similar to CVs obtained in direct measurements
(Table I).
Within-photo measurement error for the same
observer was determined by re-measuring photos on
three separate occasions. The mean CV for a total of
three photos was 0.006 for EAR-TORUS, 0.005 for
EAR-CREST, and 0.020 for ARM measurements,
respectively. When two different observers collected
measurements from the same photographs (n ¼12), the
mean CV was 0.051 (range ¼0.027–0.069) (Table II).
Measurements between observers were highly corre-
lated, for laser (R
2
¼0.9947, P<0.00001) and distance-
meter (R
2
¼0.9952, P<0.00001) techniques.
Accuracy
When comparing photogrammetric measure-
ments to the mean of direct measurements for each
target, results again demonstrated low errors
(Table I). For repeated measurements obtained using
the parallel laser technique, the average percent
error with respect to direct measurements ranged
between 2.72% and 5.20%; measurements collected
using the distance meter method were on average
2.20–7.51% different from the corresponding direct
measurements. The maximum error recorded for any
single measurement, considering all body parts, was
9.63% and 11.00% using the parallel laser and
distance meter techniques, respectively. Across all
measures, and for both techniques, the percent error
between photogrammetric and direct measurements
showed no significant relationship with observer
distance (Fig. 3).
Finally, when mean individual measurements
across structures were combined, correlations between
Am. J. Primatol.
Photogrammetry and Body Measurements in Gorillas /7
direct measurements and corresponding parallel laser
and distance meter measurements were highly signifi-
cant with both R
2
values and slopes approaching 1.0
(parallel lasers –direct measurements: R
2
¼0.9989,
P<0.0001; distance-meter –direct measurements:
R
2
¼0.9990, P<0.0001) (Fig. 4).
Results: Measurements of Static Objects
Two static objects, one large and one small, were
measured: (1) an interior office door (DOOR); and
(2) an emergency exit sign frame (SIGN). For all
laser alignments performed for these tests, the
difference between direct measurements and
photogrammetric-derived measurements of the tar-
get in both horizontal and vertical axes, respectively,
was no greater than 1.25%.
Precision
Repeated measurements of static objects collected
directly by a single observer (JG) showed uniformly
low CVs. The mean sign length was 29.870.12 cm
(CV ¼0.004), while the mean door length was
93.53 0.25 cm (CV ¼0.003). Using photogramme-
try, tests of intraobserver between-photo measure-
ment error demonstrated CVs that were similar
for both methods, for DOOR (laser CV¼0.010;
distance meter CV¼0.010) and SIGN length (laser
TABLE I. Measurements of Gorillas Collected Directly (Top), Using the Parallel Laser Method (Middle), and Using
the Distance Meter Method (Bottom), by a Single Observer (JG)
Measurement ID ND range (cm) Mean (cm) SD CV % Error: Mean (min–max)
Direct
EAR-CREST Charlie 5 n/a 23.30 0.45 0.02 n/a
Sukari 4 15.30 0.76 0.05
Kuchi 3 16.33 0.58 0.04
Lulu 3 15.33 0.29 0.02
EAR-TORUS Charlie 4 n/a 12.13 0.25 0.02 n/a
Sukari 4 11.50 1.08 0.09
Kuchi 3 10.33 0.58 0.06
Lulu 3 9.67 0.29 0.03
ARM Charlie 2 n/a 56.75 0.35 0.01 n/a
Sukari 4 40.40 0.89 0.02
Kuchi 4 46.75 1.44 0.03
Lulu 5 39.70 0.84 0.02
Parallel lasers
EAR-CREST Charlie 4 824.1–1,293.9 22.93 0.69 0.03 2.72 (1.35–5.52)
Sukari 4 838.6–1,229.8 15.88 0.49 0.03 4.33 (1.33–6.33)
Kuchi 4 894.9–1,273.7 17.08 0.76 0.04 4.81 (0.61–9.63)
Lulu 4 765.6–1,023.4 16.04 0.38 0.02 4.61 (2.00–7.40)
EAR-TORUS Charlie 4 824.1–1,293.9 12.49 0.50 0.04 3.81 (0.40–7.72)
Sukari 4 838.6–1,229.8 11.15 0.31 0.03 3.08 (0.07–6.58)
Kuchi 4 894.9–1,273.7 10.87 0.18 0.02 5.20 (3.79–8.24)
Lulu 4 765.6–1,023.4 9.67 0.46 0.05 3.59 (1.38–7.22)
ARM Charlie 4 711.6–956.0 56.31 2.82 0.05 3.64 (0.39–7.15)
Sukari 4 968.2–1,672.5 41.89 1.43 0.03 4.36 (1.66–7.98)
Kuchi 4 894.9–1,273.7 47.05 2.92 0.06 4.98 (0.85–8.90)
Lulu 4 765.6–1,364.2 40.05 1.93 0.05 3.74 (0.20–7.70)
Distance meter
EAR-CREST Charlie 4 824.1–1,293.9 22.16 0.73 0.03 4.88 (1.20–8.83)
Sukari 4 838.6–1,229.8 15.36 0.69 0.04 3.01 (0.20–6.58)
Kuchi 4 894.9–1,273.7 16.18 0.75 0.05 3.70 (1.38–7.01)
Lulu 4 765.6–1,023.4 15.12 0.72 0.05 3.50 (0.22–7.84)
EAR-TORUS Charlie 4 824.1–1,293.9 12.07 0.46 0.04 2.95 (0.25–5.73)
Sukari 4 838.6–1,229.8 10.78 0.21 0.02 6.30 (3.82–8.30)
Kuchi 4 894.9–1,273.7 10.30 0.28 0.03 2.20 (0.35–3.10)
Lulu 4 765.6–1,023.4 9.10 0.17 0.02 5.82 (3.70–7.99)
ARM Charlie 4 711.6–956.0 54.16 4.05 0.07 7.51 (3.98–11.00)
Sukari 4 968.2–1,672.5 40.06 1.39 0.03 2.66 (0.21–5.60)
Kuchi 4 894.9–1,273.7 44.55 2.16 0.05 6.07 (3.42–7.74)
Lulu 4 765.6–1,364.2 38.08 1.69 0.05 4.86 (1.94–9.74)
% Error represents the percent difference between corresponding measurements obtained directly and those using photogrammetric methods, calculated for
individual measurements as follows: % error ¼[(photogrammetric measurement mean direct measurement)/mean direct measurement]100. Here, the
Mean % Error represents the average of errors calculated for each of 4 repeated measurements per individual. D refers to object distance. SD: standard
deviation, CV: coefficients of variation.
Am. J. Primatol.
8/Galbany et al.
CV ¼0.010; distance meter CV ¼0.005) (Table III).
For within-photo measurement error, the mean CV
for a total of six photos (three for the SIGN and three
for the DOOR) was 0.0011 (range ¼0.0004–0.0019)
using the parallel laser method and 0.0012 (range
¼0.0004–0.0019) using distance meter.
Results of inter-observer error tests are also
reported in Table III. When two observers (JG, DA)
collected measurements from the same photographs,
measurements were not significantly different for
SIGN (Wilcoxon Test, z¼0.3145, P¼0.7532) and for
DOOR (Wilcoxon Test, z¼0.5345, P¼0.5930). Finally,
when images were collected by two different photog-
raphers (JG, DA) and measured by JG, the results were
not significantly different (Wilcoxon Test, z¼0.1048,
P¼0.9165; and z¼0.5241, P¼0.6002).
Accuracy
When comparing measurements obtained directly
to those obtained using photogrammetry by a single
observer (JG), the mean percent error of repeated
parallel laser measurements was 0.49% (0.15
0.11 cm) for SIGN and 0.74% (0.69 0.72 cm) for
DOOR; for distance meter measurements, the mean
percent error was 0.62% (0.18 0.16 cm) for SIGN and
0.76% (0.71 þ0.24 cm) for DOOR (Table III).
COMPARISON AND CRITIQUE
In a rare opportunity to compare measurements
obtained directly from living great apes to measure-
ments estimated remotely, we found that two
independent photogrammetric methods, employing
TABLE II. Parallel Laser and Distance Meter Photogrammetric Measurements Obtained by Two Different
Observers (JG and DA) From the Same Photographs
Charlie ARM Lulu ARM
Observer PhotoID D L DM PhotoID D L DM
JG 1 844.3 56.53 60.92 4 1,008.2 39.62 38.19
JG 2 711.6 53.48 51.95 5 1,023.4 41.23 40.47
JG 3 824.1 56.30 52.93 6 765.6 38.19 37.39
DA 1 844.3 57.60 60.62 4 1,008.2 43.31 41.35
DA 2 711.6 57.23 55.60 5 1,023.4 42.80 42.27
DA 3 824.1 57.39 54.10 6 765.6 38.22 38.03
Mean 56.42 56.02 40.56 39.62
SD 1.53 3.88 2.24 2.02
CV 0.027 0.069 0.055 0.051
Charlie EAR-TORUS Lulu EAR-TORUS
PhotoID D L DM PhotoID D L DM
JG 7 711.6 12.95 12.50 10 1,008.2 9.53 9.07
JG 8 824.1 11.94 11.43 11 1,023.4 9.43 9.31
JG 9 1,022.4 13.06 12.52 12 1,014.9 10.36 9.23
DA 7 711.6 12.57 12.39 10 1,008.2 9.74 9.63
DA 8 824.1 11.22 11.07 11 1,023.4 9.84 9.73
DA 9 1,022.4 13.10 12.95 12 1,014.9 9.73 9.62
Mean 12.47 12.14 9.77 9.43
SD 0.75 0.73 0.33 0.26
CV 0.060 0.060 0.033 0.028
Charlie EAR-CREST Lulu EAR-CREST
PhotoID D L DM PhotoID D L DM
JG 7 711.6 22.01 21.24 10 1,008.2 16.31 15.52
JG 8 824.1 22.74 21.77 11 1,023.4 16.13 15.93
JG 9 1,022.4 22.95 22.01 12 1,014.9 16.47 14.67
DA 7 711.6 21.72 21.41 10 1,008.2 16.28 16.10
DA 8 824.1 22.96 22.66 11 1,023.4 16.56 16.37
DA 9 1,022.4 19.68 19.39 12 1,014.9 15.48 15.30
Mean 22.00 21.41 16.20 15.65
SD 1.28 1.11 0.39 0.62
CV 0.058 0.052 0.024 0.039
EAR-TORUS and EAR-CREST were measured from the same photographs. D refers to object distance in centimeters. Parallel laser (L) and distance meter
(DM).
Am. J. Primatol.
Photogrammetry and Body Measurements in Gorillas /9
the use of parallel lasers [following Rothman et al.,
2008] and distance meter [following Breuer et al.,
2007], respectively, produce similar levels of accu-
racy and precision. For measurements of captive
western lowland gorillas observed at Zoo Atlanta,
mean systematic errors ranged between 2.72% and
5.20% using parallel lasers, and between 2.20% and
7.51% when using the distance meter method. The
maximum error recorded for any single measure-
ment of gorillas was 9.63% and 11.00% using the
parallel laser and distance meter techniques, respec-
tively. Moreover, direct and photogrammetric meas-
urements of gorillas were highly correlated, with R
2
values and slopes approaching 1.0. In static object
tests, mean error was 0.76% or less, across target
objects and techniques. Thus, maximum errors in our
gorilla measurements were slightly higher for the
distance meter method. However, as suggested
previously, both techniques show great promise for
research applications requiring morphological data
from primates, where inter-individual or inter-group
differences being examined are greater than mea-
surement errors reported here.
Previous photogrammetric linear measurements
of red colobus monkeys obtained using parallel lasers
showed lower errors [mean error 1.7% and maximum
error 5.0%; Rothman et al., 2008] compared to
measurements obtained directly from living gorillas
examined here. However, mean errors reported here
are similar, or lower, than those reported for
measurements obtained from howler monkeys using
the parallel laser method [for the latter, reported
mean error across all measurements ¼3.62%, mean
error for distal forelimb ¼6.3%, and for distal
hindlimb ¼13.1%; Barrickman et al., 2015]. How-
ever, accuracy of measurements of static objects
reported here are very similar, or lower, than that
described in previous studies using the same
photogrammetry techniques: 0.09% and 1.7%, re-
spectively [Breuer et al., 2007; Rothman et al., 2008].
As noted by previous investigators [Bergeron,
2007; Rothman et al., 2008], accuracy of the parallel
laser method is highly dependent upon proper
alignment of the lasers. As we have now imple-
mented this method in a field setting to study wild
gorillas [Abavandimwe et al., 2015], we have found
the apparatus to be very stable. Alignment of the
lasers is checked daily before going into the field,
according to the following criteria (based on results of
our tests above): (1) the distance between paired
vertical lasers and between paired horizontal lasers
is 4.0 0.05 cm (i.e., allows for 1.25% error in the
calibration); and (2) the difference (% error) between
Fig. 3. OLS regressions of percent error of photogrammetric
measurements on observer distance for EAR-CREST, EAR-TORUS,
and ARM length, plotted separately for laser and distance meter
(DM) methods.
Fig. 4. OLS regressions of mean individual measurements
obtained using (A) the parallel laser method and (B) the distance
meter method, against measurements obtained directly (hand
measurements).
Am. J. Primatol.
10 / Galbany et al.
TABLE III. Measurements of Static Objects Collected by a Single Observer (JG/JG) and Two Observers (JG/DA and DA/JG) Using Direct and Two
Independent Photogrammetric Methods: Parallel Laser and Distance Meter
Sign length Door length
Photogrammetry Photogrammetry
Observer:
photographer/
measurer Direct
Object
distance
Principal
distance
Laser
Msmt
Laser
error
Laser %
error
DM
msmt
DM
error
DM %
error Direct
Object
distance
Principal
distance
Laser
Msmt
Laser
error
Laser %
error
DM
msmt
DM
error
DM %
error
Intra-observer
JG/JG 29.80 723.4 202.85 29.61 0.26 0.86 29.50 0.37 1.24 93.80 202.79 92.93 0.60 0.65 94.45 0.91 0.98
JG/JG 29.80 1,037.0 202.23 29.91 0.04 0.15 29.75 0.11 0.38 93.50 1,037.8 202.22 93.51 0.02 0.02 93.09 0.45 0.48
JG/JG 30.00 1,379.5 201.54 29.73 0.14 0.46 29.79 0.07 0.24 93.30 1,323.3 201.65 94.98 1.45 1.55 92.77 0.76 0.81
Mean 29.87 29.75 0.15 0.49 29.68 0.18 0.62 93.53 93.81 0.69 0.74 93.44 0.71 0.76
SD 0.12 0.15 0.11 0.16 0.16 0.25 1.06 0.72 0.89 0.24
CV 0.004 0.010 0.005 0.003 0.010 0.010
Inter-observer
JG/DA 29.80 723.4 202.85 29.61 0.22 0.75 29.38 0.45 1.52 93.80 754.2 202.79 93.64 0.33 0.35 94.54 0.57 0.61
JG/DA 29.80 1,037.0 202.23 30.26 0.43 1.43 29.63 0.20 0.68 94.00 1,037.8 202.22 93.88 0.09 0.09 93.16 0.80 0.86
JG/DA 29.90 1,379.5 201.54 29.61 0.22 0.75 29.79 0.04 0.15 94.10 1,323.3 201.65 94.32 0.35 0.38 93.25 0.71 0.76
Mean 29.83 29.83 0.29 0.98 29.60 0.23 0.78 93.97 93.95 0.26 0.27 93.65 0.70 0.74
SD 0.06 0.38 0.12 0.21 0.21 0.15 0.34 0.15 0.77 0.12
CV 0.004 0.010 0.007 0.004 0.004 0.010
DA/JG 29.80 744.5 202.81 29.36 0.51 1.71 29.84 0.02 0.07 93.80 763.0 202.77 94.06 0.52 0.56 92.40 1.13 1.21
DA/JG 29.80 1,009.6 202.28 29.33 0.53 1.79 30.01 0.14 0.47 93.50 1,063.8 202.17 93.23 0.30 0.32 93.36 0.17 0.18
DA/JG 30.00 1,364.9 201.57 29.93 0.07 0.22 29.67 0.19 0.65 93.30 1,337.9 201.62 93.50 0.04 0.04 93.10 0.44 0.47
Mean 29.87 29.54 0.37 1.24 29.84 0.12 0.40 93.53 93.59 0.29 0.31 92.95 0.58 0.62
SD 0.12 0.34 0.26 0.17 0.09 0.25 0.42 0.24 0.50 0.50
CV 0.004 0.010 0.010 0.003 0.005 0.005
All measurements are reported in centimeters. Parallel laser (Laser) and distance meter (DM). SD, standard deviation; CV, coefficients of variation.
Am. J. Primatol.
Photogrammetry and Body Measurements in Gorillas /11
direct measurements and photogrammetric-derived
measurements of the target in both horizontal and
vertical axes, respectively, is no greater than 1.25%.
Despite carrying the apparatus over steep rugged
terrain and long treks to our study groups, we find it
necessary to realign the lasers very infrequently, on
average only once a week. Further, addition of a third
laser into the configuration offers a practical advan-
tage in that the alignment can be easily checked both
in the field and acquired photographs; if all three
lasers have remained stable in their position, they
should remain equidistant with horizontal and
vertical pairs positioned at right angles from one
another.
Although Rothman et al. [2008] reported that
mean errors between direct and photogrammetric
measurements were similar for tests of static/
inanimate objects versus colobus monkey tail
lengths, we found that accuracy in static object
measurements was higher than that calculated for
gorilla measurements. Several factors may account
for this difference. Static objects present hard
boundaries, unlike gorillas whose body surfaces are
covered with long hair; this “fuzzy boundary”
problem means that determining landmarks from
photos, and also for direct measurements, is more
challenging. Further, anatomical features having
more three-dimensionally complex morphologies
(e.g., head size or sagittal crest size) or whose
surfaces may be curved present different challenges
for reliably identifying landmarks for measurement
in photographic images, although it is noted that
direct measurements obtained across a mesh parti-
tion here are also affected by these kind of errors.
These factors may also account for differences in
percent error between the current study for gorilla
measurements and the low errors reported by
Rothman et al. [2008] for colobus tail lengths, which
as comparatively linear structures have start and
end points that may be more easily observed (e.g.,
when hanging downwards) and defined. Barrickman
et al. [2015] also reported higher mean errors
between direct and photogrammetric measurements
of fore- and hindlimb lengths in howler monkeys. For
all of these reasons, validation tests of photogram-
metric techniques that incorporate measurements
obtained directly from study animals are likely to
yield more realistic assessments of measurement
error, compared to tests using static objects. Choice of
body dimensions and species to be measured may
also contribute to differences in error, depending on
the three-dimensional complexity of those features
and the ease with which landmarks can be identified
[also see Barrickman et al., 2015].
Moreover, due to problems associated with
parallax, any deviation in the target measurement
plane away from or towards the camera can contrib-
ute to measurement error. Although only those
photos of individuals in norma lateralis orientation
were selected, minor deviations from this plane may
be difficult to detect and thus may have impacted the
results reported here. In this latter respect, Bergeron
[2007] and Rothman et al. [2008] suggested that in
those instances in which the measured surface is
tilted, the maximum measurement from a set of
measurements is probably closest to the true value.
However, in our results this was not the case; the
mean of independent measurements was closer to the
direct measurement, while the maximum of indepen-
dent measurements was higher. Linear correlations
between the direct and the mean photogrammetric
measurements, both by laser and distance meter
techniques, were very significant and presented a
slope close to 1.0. This finding may be influenced by
error in our direct body measurements, given neces-
sary challenges associated with the Zoo Atlanta
measurement protocol. Where error is random or
non-biased, using the average of a set of repeated
measurements should allow for better discrimination
between individuals [P
erez-P
erez et al., 1990].
Another source of error to consider in the parallel
laser method is that small errors in the calibration
could be magnified considerably when large objects
are measured, for example the arm. An ideal solution
to this problem would be to increase the distance
between paired lasers, thus eliminating the need to
extrapolate beyond the calibration scale. However,
given the nature of our study subjects and their target
dimensions to be measured, this is not feasible.
Finally, our results are promising for future
comparative studies across species and study sites,
where field logistics and/or the behavior of study
animals dictate the use of one technique over
another. Results produced from the two photogram-
metric methods were broadly similar when compared
to direct measurements and in our intra- and inter-
observer error tests. In fact, mean direct measure-
ments presented a very strong linear correlation with
mean measurements obtained from both photogram-
metric techniques. However, there were some minor
differences in the results generated using these two
methods. In our tests, error was often slightly higher
using the distance meter method. While our analyses
showed no relationship between observer distance
and percent error with respect to direct measure-
ments for either technique, the distance meter
method may introduce additional sources of error
that should be considered in field applications. For
instance, calculations of object length using this
method may be impacted by errors in principal
distance estimation using the prediction equation
generated above. Further, errors in distance meter
measurements may have a greater impact on object
length at short observer distances. It may also be the
case, however, that measurement accuracy may
decrease at longer distances, if the projection path
of the distance meter from observer to target is not
horizontal [Elhassan & Ali, 2011].
Am. J. Primatol.
12 / Galbany et al.
By testing the accuracy and precision of two
independent photogrammetric techniques using liv-
ing great ape subjects, our results demonstrate the
potential of these approaches for collecting morpho-
logical data at sites where subjects are observed at
close range and measurement conditions are consis-
tent with those described here. Error associated with
these methods, particularly for the parallel laser
method in which mean % error was 5.2% or less, is
broadly similar to that reported for a range of primate
morphology studies[e.g., Bailey, 2004; Galbany et al.,
2005; Kimura & Hamada, 1996; Sherwood et al., 2004;
Spoor & Zonneveld, 1995]. Thus, these techniques
may be of great value to primatologicalinvestigations,
allowing researchers to address a wide range of
previously intractable questions requiring the collec-
tion of morphological data from wild primates, and
thus show promise to contribute to research on a
variety of topics. Three-dimensional photogramme-
try, as employed in a range of other applications [e.g.,
McGlone, 2013], also represent an exciting new
direction. Particularly, 3D approaches would address
some of the above limitations of 2D photogrammetric
techniques, including the requirement that target
landmarks lie in a plane perpendicular to the optical
axis, thus expanding the range of measures that
could be accurately estimated. Though, since this
requires incorporation of multiple camera systems
collecting simultaneous images from different van-
tage points, the logistics of implementing a 3D
approach in forested habitats with high vegetation
density represents a challenge that would need to be
overcome.
ACKNOWLEDGMENTS
The authors thank Zoo Atlanta and the Primate
Staff for their support on direct measurements of the
gorillas and assistance, especially Regina Paxton
Gazes, Meg Sosnowski, Jodi Carrigan and Kristina
Krickbaum. We are also grateful to Jean Paul Hirwa,
who provided technical assistance; and Damien
Caillaud, Courtney Fitzpatrick, N
uria Garriga and
Sergio Tomey-Garc
ıa who provided helpful com-
ments, and to our anonymous reviewers whose
comments greatly improved the manuscript.
REFERENCES
Abavandimwe D, Galbany J, Breuer T, et al. 2015. Abstract:
Body growth in wild mountain gorillas (Gorilla beringei
beringei) from Volcanoes National Park, Rwanda. American
Journal of Physical Anthropology 156:65.
Abramoff MD, Magelhaes PJ, Ram SJ. 2004. Image processing
with ImageJ. Journal of Biophotonics 11:36–42.
Altmann J, Alberts SC. 2005. Growth rates in a wild primate
population: ecological influences and maternal effects.
Behavioral Ecology and Sociobiology 57:490–501.
Bailey S. 2004. A morphometric analysis of maxillary molar
crowns of Middle-Late Pleistocene hominins. Journal of
Human Evolution 47:183–198.
Barrickman NL, Schreier AL, Glander KE. 2015. Testing
parallel laser image scaling for remotely measuring
body dimensions on mantled howling monkeys (Alouatta
palliata). American Journal of Primatology 77:823–832.
Bates KT, Rarity F, Manning PL, et al. 2008. High-resolution
LiDAR and photogrammetric survey of the Fumanya
dinosaur tracksites (Catalonia): implications for the conser-
vation and interpretation of geological heritage sites.
Journal of the Geological Society 165:115–127.
Bentley J, Olson C. 2012. Field guide to lens design.
Bellingham: The International Society for Optical Engi-
neering. p 140.
Bergeron P. 2007. Parallel lasers for remote measurements of
morphological traits. Journal of Wildlife Management
71:289–292.
Bergh€
anel A, Sch€
ulke O, Ostner J. 2015. Locomotor play drives
motor skill acquisition at the expense of growth: a life
history trade-off. Science Advances 1:e1500451.
Brager S, Chong AK. 1999. An application of close range
photogrammetry in dolphin studies. The Photogrammetric
Record 16:503–517.
Breuer T, Robbins MM, Boesch C. 2007. Using photogramme-
try and color scoring to assess sexual dimorphism in wild
western gorillas (Gorilla gorilla). American Journal of
Primatology 134:369–382.
Breuer T, Breuer-Ndoundou Hockemba M, Olejniczak C,
Parnell RJ, Stokes EJ. 2009. Physical maturation,
life-history classes and age estimates of free-ranging
western gorillas-Insights from Mbeli Bai, Republic of Congo.
American Journal of Primatology 71:106–119.
Breuer T, Robbins AM, Boesch C, Robbins MM. 2012.
Phenotypic correlates of male reproductive success in
western gorillas. Journal of Human Evolution 62:466–472.
Brazeau NF, Machanda ZP, Smith AR, et al. 2013. Growth and
ecology in wild chimpanzees: a photogrammetric analysis.
82nd Annual Meeting of the American Association of
Physical Anthropologists—Undergraduate Research Sym-
posium. Knoxville, TN. Abstract: p 9–10.
Breithaupt BH, Matthews NA. 2001. Preserving paleontological
resources using photogrammetry and geographic information
systems. In: Harmon D, editor. Crossing Boundaries in Park
Management: Proceedings of the 11th Conference on Research
and Resource Management in Parks and on Public Lands.
Hancock, Michigan: The George Wright Society. p 62–70.
Butler JB, Lane SN, Chandler JH. 1998. Assessment of DEM
quality for characterizing surface roughness using close
range digital photogrammetry. Photogrammetric Record
16:271–291.
Caillaud D, Levr
ero F, Gatti S, M
enard N, Raymond M. 2008.
Influence of male morphology on male mating status and
behavior during interunit encounters in western lowland
gorillas. American Journal of Physical Anthropology 135:
379–388.
Cooper MAR, Cross PA. 1988. Statistical concepts and their
application in photogrammetry and surveying. Photogram-
metric Record 12:637–663.
Deakos MH. 2010. Paired-laser photogrammetry as a simple
and accurate system for measuring the body size of free-
ranging manta rays Manta alfredi. Aquatic Biology 10:1–10.
Deschner T, Heistermann MI, Hodges K, Boesch C. 2004.
Female sexual swelling size, timing of ovulation, and male
behavior in wild West African chimpanzees. Hormones and
Behavior 46:204–215.
Domb LG, Pagel M. 2001. Sexual swelling advertise female
quality in wild baboons. Nature 410:204–206.
Durban JW, Parsons KM. 2006. Laser-metrics of free-ranging
killer whales. Marine Mammal Science 22:735–743.
Elhassan IM, Ali AS. 2011. Comparative study of accuracy in
distance measurement using: Optical and digital levels.
Journal of King Saud University - Engineering Sciences
23:15–19.
Am. J. Primatol.
Photogrammetry and Body Measurements in Gorillas /13
Emery MA, Whiten PL. 2003. Size of sexual swelling reflects
ovarian function in chimpanzees (Pan troglodytes). Behav-
ioral Ecology and Sociobiology 54:340–351.
Falkingham PL. 2012. Acquisition of high resolution 3D
models using free, open-source, photogrammetric software.
Palaeontologia Electronica 15:15.
Fiete RD. 2013. Elements of photogrammetric optics. Chapter
4, In: McGlone JC, editor. Manual of Photogrammetry. 6th
edition. Bethesda: American Society for Photogrammetry
and Remote Sensing. p 359–450.
Fitzpatrick CL, Altmann J, Alberts SC. 2014. Sources of
variance in a female fertility signal: exaggerated estrous
swellings in a natural population of baboons. Behavioral
Ecology and Sociobiology 68:1109–1122.
Galbany J, Mart
ınez LM, L
opez-Amor HM, et al. 2005. Error
rates in dental buccal microwear quantification using
Scanning Electron Microscopy. Scanning 27:23–29.
Gavan JA, Washburn SL, Lewis PH. 1952. Photography: an
anthropometric tool. American Journal of Physical Anthro-
pology 10:331–354.
Geoghegan B. 1953. The determination of body measurements,
surface area and body volume by photography. American
Journal of Physical Anthropology 11:97–120.
Greivenkamp JE. 2004. Field guide to geometrical optics.
Bellingham: The International Society for Optical Engi-
neering. p 117.
Hugemann W. 2010. Correction lens distorsions in digital
photographs. Leverkusen, Germany: Ingenieurb€
uro Mor-
awski þHugemann. p 12.
Ireland D, Garrott RA, Rotella J. 2006. Development and
application of mass-estimation method for Weddell seals.
Marine Mammal Science 22:361–378.
Jaquet N. 2006. A simple photogrammetric technique to
measure sperm whales at sea. Marine Mammal Science
22:862–879.
Johnson SE. 2003. Life history and the competitive environ-
ment: trajectories of growth, maturation, and reproductive
output among chacma baboons. American Journal of
Physical Anthropology 120:83–98.
Kerr DC. 2004. Lens principal and nodal points. The Pumpkin
3:1–5.
Kimura T, Hamada Y. 1996. Growth of wild and laboratory
born chimpanzees. Primates 37:237–251.
Kingslake R. 1992. Optics in Photography. Bellingham: The
International Society for Optical Engineering. p 289.
Kurita H, Suzumura T, Kanchi F, Hamada Y. 2012.
A photogrammetric method to evaluate nutritional status
without capture in habituated free-ranging Japanese
macaques (Macaca fuscata): a pilot study. Primates 53:7–11.
Lu A, Stinespring-Harris A, McCann C, Beehner JC. 2013.
Using digital photogrammetry to estimate growth in wild
geladas. American Journal of Physical Anthropology
150:182–183.
Machanda Z, Brazeau NF, Castillo E, et al. 2015. Abstract:
musculoskeletal growth patterns in wild chimpanzees (Pan
troglodytes). American Journal of Physical Anthropology
156:210.
Mallison H, Wings O. 2014. Photogrammetry in paleontology—a
practical guide. Journal of Paleontological Techniques
12:1–31.
McGlone JC, editor. 2013. Manual of Photogrammetry.
Bethesda: American Society for Photogrammetry and
Remote Sensing. p 1318.
Mikhail E, Bethel J, McGlone JC. 2001. Introduction to
modern photogrammetry. New York: John Wiley & Sons. p
479.
Morgan BJ, Lee PC. 2003. Forest elephant (Loxodonta
africada cyclotis) stature in the R
eserve de Faune du
Petit Loango, Gabon. Journal of Zoology, London 259:337–
344.
P
erez-P
erez A, Alesan A, Roca L. 1990. Measurement error:
inter- and intraobserver variability. An empiric study.
International Journal of Anthropology 5:129–135.
Pusey AE, Oehlert GW, Williams JM, Goodall J. 2005.
Influence of ecological and social factors on body mass of
wild chimpanzees. International Journal of Primatology
26:3–31.
Rothman JM, Chapman CA, Twinomugisha D, et al. 2008.
Measuring physical traits of primates remotely: the use of
parallel lasers. American Journal of Primatology 70:1–5.
Sherwood CC, Cranfield MR, Mehlman PT, et al. 2004. Brain
structure variation in great apes, with attention to the
mountain gorilla (Gorilla beringei beringei). American
Journal of Primatology 63:149–164.
Shrader AM, Ferreira SM, van Aarde RJ. 2006. Digital
photogrammetry and laser rangefinder techniques to
measure African elephants. South African Journal of
Wildlife Research 36:1–7.
Spoor F, Zonneveld F. 1995. Morphometry of the primate bony
labyrinth: a new method based on high-resolution computed
tomography. Journal of Anatomy 186:271–286.
Tanner JM, Weiner JS. 1949. The reliability of the photo-
grammetric method of anthropometry, with a description of
a miniature camera technique. American Journal of Physi-
cal Anthropology 7:145–186.
Tubbs G, Ito K. 2002. HD Lens optical design relating to focus
breathing and depth of field issues. SMPTE Fall Conference,
Pasadena, 2002.
Walker S, AlspaughD. 2013. A brief history of photogrammetry.
In: Chris McGlone J, editor. Manual of Photogrammetry. 6th
edition. Bethesda: American Society for Photogrammetry
and Remote Sensing. p 1–62.
Am. J. Primatol.
14 / Galbany et al.
