You are on page 1of 19

The Photogrammetric Record 24(125): 5–22 (March 2009)

A PHOTOGRAMMETRIC SURVEYING METHOD FOR


FIELD APPLICATIONS
Dirk H. Rieke-Zapp (zapp@geo.unibe.ch)
Ralph Rosenbauer (rosenbauer@arch.unibe.ch)
Fritz Schlunegger (fritz.schlunegger@geo.unibe.ch)
University of Bern, Switzerland

Abstract
A photogrammetric method is presented for mapping rock outcrops and other
objects in the field. Special attention was paid to simplifying the workflow and to
minimising extra bulk and weight of the required equipment while maximising the
accuracy of the survey. The minimum equipment needed for the surveys added less
than 0Æ5 kg to the backpack of a field scientist, or as little as 0Æ2 kg assuming that a
suitable camera was already part of the typical equipment carried in the field. Data
acquisition in the field took less than 15 min, while image orientation and preparation
for stereoviewing took less than 30 min even for a user with little training. Best results
were accomplished taking two convergent images as well as three images parallel to
the object of interest. Two test sites were surveyed with the method, covering volumes
of 3Æ5 · 5 · 3 m3 and 18 · 20 · 5 m3, to identify the most accurate adjustment method.
The maximum length measurement error (LME) was calculated for 28 and 78
distances in the smaller and larger volume, respectively, based on a comparison of
the photogrammetric survey with an independent total station survey of the same
signalised points. The maximum LME ranged between 0Æ005 and 0Æ002 m for the first
test site where three cameras were tested, and was 0Æ021 m for the larger test site
where only one camera was tested. The rmse values of the LMEs ranged between
0Æ003 and 0Æ001 m for the first and 0Æ010 m for the second test site, respectively. The
smallest and lightest camera, a Ricoh GR Digital, yielded the most accurate results in
object space when interior orientation was calibrated on the job. The Sigma SD14
did not require on-the-job calibration for accurate results and was the camera with
the best geometric stability. The third camera evaluated, a Rollei d7 metric5, also
yielded good results, but could not deliver the extra value that would be expected
from a metric camera designed for photogrammetric surveys.

Keywords: camera calibration, digital camera, field survey, geology,


geomorphology, photogrammetry

Introduction
Earth scientists require data for documentation and quantification of their fieldwork.
While natural phenomena can be drawn on a map or documented with photographs,

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
Blackwell Publishing Ltd. 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street Malden, MA 02148, USA.
Rieke-Zapp et al. A photogrammetric surveying method for field applications

quantification requires application of surveying methods that typically involve a fair amount of
extra equipment on location as well as user expertise on how to operate this equipment. At the
same time, the main goal of fieldwork is usually not the most precise measurement, but the
description of the local geology or geomorphology of a given area. Often the need for
measurements arises in the field when surveying equipment is not available because it would
be too heavy or too delicate to take to the field just in case it is needed. The standard backpack
of a typical field scientist does not include surveying equipment other than a ruler, a yardstick
or a compass as well as a camera; the latter is often not recognised as a surveying tool. The
extra bulk and weight of more elaborate surveying equipment often rules out its use in remote
areas as well as for air travel to field sites abroad.
There is thus a need for lightweight surveying equipment customised for use by earth
scientists in the field. Earth scientists take a camera to the field because the imagery serves as a
permanent record to aid their memory in qualitative analysis on return to the office. What is often
not realised is that oriented imagery with stereoscopic overlap can be employed to gather metric
information of an object for quantitative work as well: in short, images can be used both for
documentation and measurement. Photogrammetry, and digital photogrammetry in particular,
may thus compensate for the absence of a lightweight field surveying method. Photogrammetry
has been applied previously to provide accurate field data for projects in geomorphology and
geology alike (Meydenbauer, 1912; Pike, 2000) and is hence far from being a completely new
technique for earth scientists. Photogrammetry is the process of deriving metric information about
an object through measurements made on photographs of the same object (Mikhail et al., 2001).
Photogrammetric surveys yield high precision results. Precision depends mostly on image scale,
the configuration of images, the precision of point measurements in the images, the precision of
ground control points and the stability of the camera geometry. While any camera may be used for
photogrammetric surveys, best results will be accomplished with cameras that produce quality
images and employ precise alignment of the photographic lens–sensor system (Rieke-Zapp et al.,
2005; Luhmann et al., 2007; Wackrow et al., 2007). Since a digital camera is already a standard
tool for field scientists nowadays, it will not add extra weight or bulk to the backpack. This implies
that photogrammetry would be the perfect candidate for a suitable field surveying method. The
measurement of oriented images with stereoscopic overlap is a standard task performed with
photogrammetric workstations that are often readily available in universities and research
institutions in conjunction with geographic information systems and/or remote sensing
software.
Although there is a long list of publications on applied photogrammetry for projects in the
earth sciences (see Pike, 2000 for a review), two main reasons can be identified that have so far
kept field scientists from a more extensive application of photogrammetry. Firstly, for retrieval
of object coordinates from images, a coordinate system has to be established which is typically
accomplished by a precision survey of signalised points employing a total station or a geodetic
GNSS—just the tools that were precluded above owing to their size and weight or cost and
availability. In most projects where photogrammetry was used for quantification, the survey
was located in an easy access area (Welch and Jordan, 1983), or even inside a laboratory
(Lascelles et al., 2002; Rieke-Zapp and Nearing, 2005), and the projects were designed to suit
the equipment. A field method should work the other way around, with a method suitable for
the project encountered in the field. Secondly, potential users have been ignorant of
photogrammetry or have had reservations about its potential. While digital cameras are almost
ubiquitous and digital photogrammetric workstations for image measurement can be found as
add-ons to standard software packages or may even be readily available in research institutions,
the potential of these tools lies idle or is unrecognised as many field scientists lack training in
photogrammetry or consider its field use to be too complicated.

6  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

The objectives of this contribution are to present a photogrammetric method that (a) adds
minimum bulk and weight to a field scientist’s backpack, (b) is as simple as possible, yielding
results as accurate as possible and (c) is based on well-established photogrammetric methods to
allow scaling according to the demands of the project.

Materials and Methods


Cameras Evaluated for the Project
Three different digital cameras were evaluated (Table I). They represent a range of
differently sized cameras that were assumed to be suitable for photogrammetric measurements.
The Rollei d7 metric5 is a single lens reflex (SLR) camera with an integrated lens and was
introduced in 2002. The camera is sold for photogrammetric applications and shipped with a
calibration certificate listing the parameters of the interior orientation of the camera as well as
additional lens distortion parameters required to define the interior geometry of the camera.
Focal length and distance setting of this camera were fixed to retain the calibrated parameters
of the interior orientation. The camera was calibrated on a testfield before it was used in the
field to refresh the outdated calibration by the manufacturer. The second camera was a Ricoh
GR Digital (Table I), a compact digital camera introduced in 2005. Unlike the Rollei camera,
the Ricoh was not designed for photogrammetric applications. It is equipped with a wide-angle
lens that can be focused manually for different distance settings. The parameters of interior
orientation as well as distortion parameters were calibrated on a testfield before the camera was
used in the field (Table I). The parameters apply only for the distance setting that was used for
calibration. The retracting lens mechanism did not allow taping or fixing the calibrated distance
setting. Therefore, the manual focus setting of the camera was applied and the camera was used
with the same distance setting in the field that was calibrated on the testfield. The third camera
was a Sigma SD14 with a Sigma 24 mm screw-mount (M42) wide-angle lens fitted via an
adapter to the camera (Table I). The camera was introduced in 2007 while the lens design dates
back more than 20 years. The screw-mount lens was a manual focus lens and was preferred

Table I. Specifications of cameras evaluated.


Camera Rollei d7 Ricoh GR Sigma SD14 with Sigma Super-Wide II
metric5 digital 2Æ8/24 mm(1)
Weight (ready to go) [kg] 0Æ750 0Æ200 0Æ780
Field of view (horiz./vert.) [] 62/48 64/50 47/32
Pixel pitch [mm] 0Æ0035 0Æ0022 0Æ00785
Pixel number (rows/columns) 2552/1920 3264/2448 2640/1760
Test site Landiswil Landiswil Landiswil Illgraben
Calibrated focal length [mm] 7Æ452 5Æ816 24Æ48 24Æ36623
Principal point (x/y) [mm] 0Æ1736/ 0Æ01418/)0Æ0931 )0Æ0608/0Æ0294 )0Æ06777/0Æ05055
0Æ0086
Additional parameters )2Æ26E-3/ )1Æ458E-3/ )2Æ329E-4/ )2Æ31565E-4/
(A1/A2/r0) [mm] 4Æ198E-5/ 3Æ194E-5/ 3Æ525E-7/ 3Æ47406E-7/
3Æ00 3Æ00 8Æ00 8Æ00
Calibrated distance setting 3m 3 m(2) 3m Infinity
Aperture value applied (f/) 5Æ6 7Æ1 9Æ5 9Æ5
Exposure time, handheld [s] 245)1 125)1 60)1 125)1
ISO rating applied 100 64 50 200
(1)
The M42 screw-mount lens was fitted via adapter to the Sigma SD lens mount.
(2)
‘‘Snapshot’’ setting of the camera.

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 7
Rieke-Zapp et al. A photogrammetric surveying method for field applications

over current autofocus lenses because calibration tests with the screw-mount lens yielded much
better geometric stability. The distance setting of the lens was fixed with tape and the lens was
kept on the camera after calibration. The camera was also calibrated on a testfield before field
use (Table I). All three cameras were used on the Landiswil test site. Images of the Illgraben
test site were only taken with the Sigma camera. Because the two tests were undertaken several
months apart and the lens on the Sigma was unmounted during this time, different calibrations
of the Sigma camera apply for the two test sites (Table I).

Field Set-up
The definition of a three-dimensional object coordinate system was based on the 3–2–1
method (Luhmann et al., 2007). Defining six coordinate values for three points was sufficient
to fix an arbitrary Cartesian coordinate system in object space (Fig. 1). The three points fixing
the orientation of the axes were not signalised in the field and were only measured in the
images. The direction of the Z axis (pointing towards the camera) was set by definition of a
relatively left and a relatively right image position regarding the exposure stations’
X coordinate. To enable measurements in this arbitrary coordinate system, some metric
information—a scale—had to be introduced. Although a single distance would have sufficed in
this set-up, measuring three distances increased redundancy and thus the reliability of the
results without adding too much extra time to the field set-up.
Signalised control points were ping-pong balls and wooden spheres for the Landiswil and
Illgraben test sites, respectively. Ping-pong balls are well-rounded spheres with a diameter of
40 mm according to International Table Tennis Federation (ITTF) regulations (ITTF, 2007).
Spheres had the advantage that they served as omni-directional target points. In images they were
reproduced as circles regardless of the viewing angle and on the target rock outcrop wall the
distance could be measured pointing from any direction. While ping-pong balls weighed
0Æ0027 kg (ITTF, 2007), wooden spheres weighed more—approximately 0Æ020 to 0Æ030 kg—but

Fig. 1. System fixing with the 3–2–1 technique and scale information from distance measurements.
Screenshot from Rollei CDW software ‘‘system-fixing’’ window.

8  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

had the advantage that they were more rugged and were available in different sizes allowing
easy adjustment of sphere diameter to image scale. Orange ping-pong balls or wooden spheres
painted in pink were easily recognised in the field and in the images. The spheres were fixed
with modelling clay on the outcrop to stay in place while measurements and images were
taken. These spherical control points were the only signalised points in the field.

Distance Measurement
The distance measurements between the spheres were performed with a Leica Disto A3
laser distance meter. The measurement accuracy (1 r) of this device was 0Æ003 m for a distance
of up to 30 m (Leica, 2006); weight including battery was 0Æ156 kg. Handheld, the Leica A3
laser dot could be pointed precisely towards the centre of a ping-pong ball for up to 7 m; when
the hand was supported by a plastic bag filled with sand, this distance was increased to more
than 20 m. The laser distance meter was placed tangentially to one sphere while aiming
towards the centre of the targeted sphere. Adding the diameter of a sphere to the measured
distance represented the distance between the centre points of the two spheres, given that
spheres with the same diameter were used. Measurements were taken in continuous mode of
the laser distance meter. Typically, the range of all accepted distances was within 0Æ001 m.
Visibility of the laser dot in daylight was limited to a maximum distance of approximately 6 m
and was greatly improved, to 20 m and beyond, when wearing the red goggles available as an
accessory.
While the operation of the Leica A3 was straightforward, it was found that the distance
reading of the laser beam was affected by the reflectance of the target material. For the wooden
spheres no deviation was found from the measured distance displayed on the device compared
to the actual distance as measured with a steel tape. However, when pointing at ping-pong balls
the Leica A3 readings were not accurate for distances up to approximately 6Æ7 m. Based on 10
readings taken for distances between 0Æ3 and 8Æ0 m the following error function was computed
by simple linear regression:

D ¼ 00015X þ 96785 ð1Þ

where X was the distance reading displayed on the laser distance meter and D was the
deviation of the distance reading in millimetres that had to be subtracted from X to yield the
corrected distance. The R2 of the function was 0Æ99. For a distance of 3000 mm D becomes
5 mm and must be compensated to achieve accurate results.

Software Workflow
The whole workflow was optimised for image orientation in Rollei Close Range Digital
Workstation (CDW) (version 2.1; Rollei, 2002) and for image measurement in Leica
Photogrammetric Suite (LPS) (version 9.1; Leica Geosystems, 2005) software. Both products
were readily available and complemented each other well. A similar workflow may be defined
for comparable software products. Parameters of interior and exterior orientation were
transferred between the two software products. Care had to be taken to ensure that all data
referred to the same units, same parameters, and to the same rotation sequence. Rollei CDW
allows orientation of images by means of the 3–2–1 method and scale information. This
functionality was not available in LPS. LPS in turn offered stereoscopic viewing and
measurement capabilities, such as image matching and orthophoto generation, which were not
available in Rollei CDW. Stereoviewing and retrieval of three-dimensional information is part

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 9
Rieke-Zapp et al. A photogrammetric surveying method for field applications

of the standard image measurement process in LPS and is not covered here. The important
steps documented for the method were image orientation in Rollei CDW and the correct import
of orientation parameters into LPS in order to start the standard measurement process.
Some project parameters were defined in Rollei CDW in such a way as to ease subsequent
processing in LPS. The aerial rotation sequence was chosen in Rollei CDW. The rotation
angles were calculated as sequential rotations around the X, Y and Z axes as defined by the
angles x, / and j. The Z axis was pointing towards the camera by definition. Angular units
were defined in degrees, distance units in metres. Choosing these parameters limited
adjustments in the LPS import dialogue to a minimum, as they represented the default values
for the block property set-up in LPS. Stereoviewing in LPS was only possible when the Z axis
was pointing towards the camera.
Before starting fieldwork, the interior orientation (IOR) of the cameras was calibrated on a
testfield in Rollei CDW software. Only the interior orientation and radial symmetric lens
distortion parameters (Table I) were calibrated since LPS did not allow import of any further
parameters in a camera file. Calibration results from Rollei CDW required some alteration
before they could be applied in an LPS camera file. Calibrated focal length was a negative
number in Rollei CDW while it was a positive number in LPS. Simply changing the algebraic
sign resolved this difference. Rollei CDW employed the balanced model to account for radial
symmetric lens distortion with the parameters A1, A2 and r0 (Fryer, 1996). The parameters were
transferred to the K-parameter series (K0, K1, K2) employed in LPS as follows:
K0 ¼ ðA1  r02 þ A2  r04 Þ ð2Þ

K 1 ¼ A1 ð3Þ

K2 ¼ A2 : ð4Þ
All image measurements were done in Rollei CDW using the manual measurement mode.
The user had to measure three different types of points in the images. First of all the three
spheres were measured and the distances between them used in the ‘‘system fixing’’. Distances
were introduced as fixed distances; assignment of a standard error reflecting measurement
accuracy had no effect on the calculated results and was therefore omitted to ease the workflow.
In the next step, points for the definition of the coordinate axes were measured and values for
X, Y, Z coordinates were defined for the system fixing in Rollei CDW according to the 3–2–1
method (Fig. 1).
Careless selection of the points resulted in a rotated coordinate system that appeared
unnatural and made three-dimensional measurement in the stereo viewer difficult. In addition
to the points used for system and scale definition, at least six tie points were measured,
meaning that a total of at least 12 different points were measured in the images. Points were
chosen such as to cover the whole area of interest on the object in a well-distributed pattern and
to make sure that they could be identified in as many images as possible to increase redundancy
of the measurements. A total of six tie points was considered the bare minimum; adding more
points should improve the geometric coherence of the adjustment. Since the number of images
was also kept to a minimum, it was important to create sufficient overlap in the images to
ensure that points were visible in as many images as possible. Tie points should be visible in at
least three images while scale and system set-up points should be visible in more than three
images if possible. The precision of manual point measurements was assumed to be between 1
and 2 pixels for the three cameras. As a consequence the a posteriori standard deviation of the
bundle adjustment was set to 0Æ005 mm for all adjustments. Working with a small number of

10  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

images and points eased the workflow, but reduced the redundancy of the subsequent
adjustment at the same time. This made it difficult to identify errors committed during set-up or
point measurement. Some practice was required by the user to minimise the number of errors
committed and to correct errors when they were identified during the adjustment.
Rollei CDW offered a two-step approach for calculation of image orientation and object
coordinates. Multi-image orientation calculated a combined resection for all images based on
image measurements, system definition and camera calibration parameters. Results of these
calculations provided start values for the bundle adjustment, which is typically considered the
most rigorous and exacting adjustment (Mikhail et al., 2001). Both calculations yielded image
orientation and coordinates for all points in object space. The bundle adjustment can also be
used for a simultaneous calibration of the interior orientation of the camera including additional
parameters. Simple editing and renaming of Rollei CDW files allowed for direct transfer of
results in a new LPS project.

Field Tests
Two tests were performed under field conditions for an assessment of the accuracy in
object space of the method. In the first test, all three cameras were compared to evaluate the
best image set-up and orientation procedure for each camera in Rollei CDW software. For the
second test only the Sigma SD14 was available. In this test the effect of extrapolation errors on
a large outcrop wall was evaluated when the reference points covered only a small part of the
area under investigation. In this test the method was pushed to the limits for the cameras tested,
as the large size of the outcrop resulted in a per pixel resolution just sufficient to differentiate
fine material from pebbles and rocks in the images.
The first outcrop had dimensions of 3Æ5 · 5 · 3 m3, for height, width and depth,
respectively. This outcrop represented the typical volume that may be investigated in the field
for mapping details at the millimetre to centimetre scale. Three ping-pong balls were placed on
the outcrop wall (Fig. 2). In addition to that, eight target points for measurement with a total
station were placed on the wall in a well-distributed pattern (Fig. 2). The X, Y, Z coordinates of
the eight target points were determined in a total station survey to an accuracy of 0Æ002 m in all
dimensions. Between the eight points 28 distances can be calculated which were compared
with the distances calculated from the photogrammetric process. The photogrammetric
measurement was only based on the three distances measured between the ping-pong balls. No
information from the total station survey was used in the photogrammetric adjustment. The
total station targets were measured in the images and were introduced as tie points in the
photogrammetric adjustment to calculate X, Y, Z coordinates. The difference of the Euclidean
distance between the points as calculated from the photogrammetric and the total station survey
represented the length measurement error (LME) which was calculated along with the root
mean square error (rmse) of all length measurements (Table II).
This procedure was chosen in reference to VDI/VDE 2634 Part 1 (2002), a German
standard for evaluation of object-space accuracy of three-dimensional point measurement
systems based on the LME. Results were calculated for different photogrammetric adjustment
methods and for two set-ups of image acquisition (Table II).
Previous tests had shown that at least five images should be taken in the field to ensure
sufficient redundancy in the adjustment as well as for stereoviewing in LPS. A first set of
images was acquired for each camera where all images were taken in portrait format (normal
set-up). The second set of images was taken rotating the camera around the optical axis by 90
degrees after each image (rotated set-up). Rotation around the optical axis will minimise
correlation of parameters of interior orientation in cases where the five images will be used for

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 11
Rieke-Zapp et al. A photogrammetric surveying method for field applications

Fig. 2. Outcrop area in Landiswil—first test site. Ping-pong balls are marked by red dots, control points are
marked in yellow. Distances measured with the laser distance meter are shown as red dashed lines. The yellow lines
represent the 28 distances between the control points that were calculated.

Table II. Accuracy in object space for the three cameras at the Landiswil test site.
Adjustment method Accuracy in object Rollei d7 metric5(1) Sigma SD14 Ricoh GR Digital
space [mm] (2) (3)
Normal Rotated Normal Rotated Normal Rotated
Multi-image resection Max. LME 0Æ0025 0Æ0050 0Æ0026 0Æ0113 0Æ0288 0Æ0319
Rmse 0Æ0018 0Æ0025 0Æ0013 0Æ0042 0Æ0125 0Æ0173
Bundle adjustment Max. LME 0Æ0051 0Æ0083 0Æ0040 0Æ0037 0Æ0276 0Æ0274
Rmse 0Æ0026 0Æ0033 0Æ0019 0Æ0020 0Æ0127 0Æ0172
Bundle adjustment + IOR Max. LME 0Æ0027 0Æ0099 0Æ0057 0Æ0032 0Æ0048 0Æ0029
Rmse 0Æ0024 0Æ0035 0Æ0023 0Æ0019 0Æ0029 0Æ0014
Bundle adjustment + A1, A2 Max. LME 0Æ0050 0Æ0076 0Æ0037 0Æ0036 0Æ0272 0Æ0308
Rmse 0Æ0024 0Æ0029 0Æ0019 0Æ0019 0Æ0126 0Æ0171
Bundle adjustment Max. LME 0Æ0102 0Æ0101 0Æ0069 0Æ0048 0Æ0087 0Æ0052
+ IOR + A1, A2 Rmse 0Æ0056 0Æ0039 0Æ0028 0Æ0026 0Æ0041 0Æ0027
(1)
In the images taken in the normal set-up, one point was visible in only one image, reducing the longest distance to
just 4Æ6 m and the total number of distances for calculation to 21.
(2)
Five images taken in landscape format without camera rotation.
(3)
Five images taken by rotating the camera by 90 degrees around the optical axis after each shot.

an on-the-job calibration (Wester-Ebbinghaus, 1983; 1985). Taking two converging images,


one from the far left and the other from the far right (Fig. 3) of the object of interest increased
the geometric stability of the five-image set-up and eased orientation in Rollei CDW software.
The remaining three images were taken with the camera parallel to the object of interest,
the outcrop wall in this case, for easy stereoviewing in LPS. Convergent images are difficult for
stereoviewing in LPS as the software offers no function for calculation of normalised image
pairs (removal of the y parallax from an image pair).
Calibration on the job was thought to be especially helpful, if not necessary, for the two
non-metric cameras to yield accurate results in object space. Different orientation methods in
CDW were evaluated to check for best accuracy in object space, but also to test the robustness

12  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

+106
+20
+107
+102 +101
+30

+10
+105

5
+103 +104 +108

4
1 3
2

Fig. 3. Screenshot from Rollei CDW software showing camera positions (1 to 5), the ping-pong balls (10 to 30)
and control points (101 to 108). Images 1 and 5 are converging while images 2 to 4 were taken with the sensor plane
parallel to the outcrop wall.

of a given solution. Multi-image resection was used to provide starting values for a bundle
adjustment. Four different types of bundle adjustment, three including self-calibration of the
interior geometry of the camera, were compared (Table II). It should be noted that Rollei CDW
software allows calibration for the parameters A1 and A2 in combination only and that LPS did
not allow the direct import of further parameters that may be calibrated in CDW software.
A second test was performed in the Illgraben area in the Valais with the Sigma SD14
camera (Table I). Dimensions of the outcrop were 18 · 20 · 5 m3 in height, width and depth,
respectively (Fig. 4). Three wooden spheres were placed in the lower part of the outcrop that
was accessible without climbing. Again, reference points for a total station survey were placed
all over the outcrop area (Fig. 4). A total of 13 points were distributed and measured with a
total station to an accuracy of 0Æ002 m in all dimensions. No information from the total station
survey was introduced into the photogrammetric measurement process. The 13 points allowed
calculation of 78 distances for calculation of the maximum LME as well as the rmse of all
distances (Table III).

Results and Discussion


The longest distance between points of the Landiswil outcrop (Fig. 2) was approximately
5Æ7 m. Results of the Rollei d7 metric5 with all images taken in landscape format were omitted
from the discussion here because one point was only available in a single image so that the
largest distance in the volume was approximately 4Æ6 m. All cameras can reach a maximum
LME of 0Æ0050 m and less, and an rmse of 0Æ0025 m. The maximum LME accomplished with
all cameras surpassed the centimetre accuracy desired for the mapping of this outcrop wall. The
procedure of image orientation with or without calibration on the job, as well as the rotation of
images, had considerable influence on the results and showed significant differences depending
on the camera under investigation.
The Rollei camera reached best accuracy values with multi-image resection. Taking five
images, while rotating the camera by 90 degrees around the optical axis after each shot,
resulted in a maximum LME of 0Æ0050 m and an rmse value of 0Æ0025 m. The result was not
improved in the bundle adjustment, which was surprising as the bundle adjustment is typically

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 13
Rieke-Zapp et al. A photogrammetric surveying method for field applications

Fig. 4. Debris flow deposits at the Illgraben test site. Wooden spheres are marked by red dots, control points are
marked in yellow. Distances measured with the laser distance meter are shown as red dashed lines. The yellow lines
represent the 78 distances between the control points that were calculated.

Table III. Accuracy in object space of the Sigma SD14 camera at the Illgraben test site.
Adjustment method Accuracy in object space [mm] Sigma SD14
Multi-image resection Max. LME 0Æ0301
Rmse 0Æ0107
Bundle adjustment Max. LME 0Æ0429
Rmse 0Æ0169
Bundle adjustment + IOR Max. LME 0Æ0209
Rmse 0Æ0094
Bundle adjustment + A1, A2 Max. LME 0Æ0410
Rmse 0Æ0158
Bundle adjustment + IOR + A1, A2 Max. LME 0Æ0212
Rmse 0Æ0097

employed to refine and optimise results. As no blunder was detected in the data-set, it was not
clear why the combined resection in space yielded better results than the bundle adjustment.
Plotting the LME for each distance (Fig. 5) revealed that LME increased with increasing

14  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

distances, indicating a scale problem. This scale error could have been caused either by an
inaccurate interior orientation of the camera or by a scaling error in the object-space
coordinates. The same effect was visible for the other two cameras (Fig. 5), but was much less
pronounced. Since the accuracy of the total station survey and the distances measured with the
laser distance meter were of similar magnitude, a small bias of the LME, as seen for the other
two cameras, appeared conceivable. The effect appeared somewhat amplified by the Rollei
camera. On-the-job calibration of the interior orientation of the Rollei camera did not improve
the LME, indicating that the number of images or the arrangement of the images was not
sufficient to significantly improve upon the laboratory calibration. This problem was not
prominent in previous experiments where the Rollei camera showed good geometric stability.
The results of the bundle adjustment did not fall far behind the results of multi-image resection
and yielded a maximum LME of less than 0Æ010 m.
Image quality and handling of the Rollei was not on a par with the other two cameras,
showing the progress of camera and image technology over the 5 years following introduction
of the Rollei camera. Image quality became acceptable when the lens was stopped down to at
least f/5.6. The camera locked up between exposures for more than 10 s while the image was
written to the memory card. Six AA batteries were sufficient to take and briefly review 50
images.
The Sigma SD14 camera was bulkier, but not much heavier, than the Rollei and was
considered a semi-metric camera. However, application of the laboratory calibration yielded
results even better than for the Rollei metric camera. Both cameras were calibrated
approximately 2 weeks before the test. Results of the Sigma camera were similar for all

10
Length measurement error (mm)

–5

Ricoh GR digital
Rollei d7 metric5
Sigma SD14
–10
1000 2000 3000 4000 5000 6000
Distance (mm)

Fig. 5. Length measurement errors (LMEs) for all three cameras at the Landiswil test site. Results for the Rollei
and Sigma cameras are shown for the bundle adjustment of rotated images. Results for the Ricoh camera are shown
for the bundle adjustment with calibration of the interior orientation of rotated images.

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 15
Rieke-Zapp et al. A photogrammetric surveying method for field applications

adjustment methods, indicating that it was geometrically stable and that the five images
represented a robust data-set for this camera, even when calibration was performed on the job.
Calibration of the interior orientation yielded better results for the data-set with rotated images,
compared to images taken in normal orientation only. The more favourable maximum LME
values were therefore accomplished for the image series with rotated images (Table II). In the
case of the bundle adjustment, maximum LME was 0Æ0037 m and the rmse was 0Æ0020 m.
On-the-job calibration of the interior orientation slightly improved maximum LME and rmse
values to 0Æ0032 and 0Æ0019 m, respectively, indicating at the same time that on-the-job
calibration did neither significantly improve nor deteriorate accuracy in object space. As for the
Rollei camera, maximum LME revealed a trend that it would increase with increasing distance
(Fig. 5). For the Sigma camera this trend was less pronounced as the two largest distances were
reproduced fairly accurately. Even with additional parameters present in the bundle adjustment,
maximum LME and rmse values were more favourable than any result accomplished with the
Rollei camera.
The Sigma SD14 possessed the largest imaging sensor and produced the best image
quality of all three cameras. The in-camera memory buffer allowed for continuous image
acquisition in the field and the battery life was sufficient to take and review hundreds of images
under field conditions.
The Ricoh GR Digital camera showed a two-sided behaviour. Image orientation based on
the laboratory calibration yielded results worse than for the other two cameras, and far below
the desired centimetre accuracy. This clearly indicated that the geometric stability of the
camera was weak, which may be explained by the retracting lens moving in and out of the
camera when the power was switched off and on. Calibrating the camera on the job improved
accuracy in object space almost by an order of magnitude, and the camera then easily surpassed
the results accomplished with the other two cameras. Especially for the set-up with rotated
images, calibration of the interior orientation from only five images yielded excellent results. In
this case maximum LME was 0Æ0029 m and rmse was 0Æ0014 m. Calibration of the distortion
parameters (A1, A2) in the bundle adjustment did not improve the accuracy in object space.
Although the Ricoh camera surpassed the accuracy of the other two candidates after on-the-job
calibration, it is important to note that this procedure was considered a must for this camera,
and that the five-image set-up should be considered the bare minimum for successful bundle
adjustment with self-calibration of the interior orientation. The accuracy in object space
accomplished with the Ricoh camera was especially pleasing as it was by far the lightest and
smallest camera in the test.
Best image results were accomplished taking raw imagery instead of in-camera JPEG
images. In raw mode the camera froze for 5 s before the next shot could be taken. Battery life
was not an issue under field conditions. Image quality of the Ricoh GR Digital camera with 8
megapixels was below the quality of the Sigma SD14 with less than 5 megapixels. The camera
has a smaller sensor than the Sigma as well as a smaller pixel pitch that probably caused the
slightly lower image quality. The Ricoh GR Digital and the Sigma SD14 produced images with
a quality much better than that of the Rollei camera.
Results of all three cameras revealed that a five-image set-up was sufficient to accomplish
accuracy in object space that was sufficient for the intended purpose. The combination of two
convergent images with three images taken parallel to the outcrop created no problem for
image orientation in Rollei software. Rotation of images improved accuracy in object space.
The accuracy in object space was slightly reduced when all parameters (IOR, A1, A2) were
introduced as unknowns in the bundle adjustment, indicating that the solution may become
overparameterised when this is done. Calibrating the parameters of interior orientation resulted
in best accuracy in object space, while calibration of radial symmetric lens distortion had little

16  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

or no positive effect on accuracy. This indicated that unstable camera geometry almost solely
affected the position of the principal point and the focal length rather than radial symmetric
lens distortion. That said, the solutions of the bundle adjustments were geometrically stable and
the resulting accuracy was well within the limits defined for the project. While calibration of
the interior orientation was mandatory for the Ricoh GR Digital to optimise accuracy in object
space, on-the-job calibration is not necessary for the other two cameras. The Sigma SD14
showed no need for calibration and results of the Rollei camera were not improved by
calibration on the job. Considering that the accuracy in object space of the total station survey
was 0Æ002 m in all dimensions for each point, maximum LME values for distance
measurements between 0Æ0050 and 0Æ0029 m and rmse values between 0Æ0025 and 0Æ0014 m
for the three cameras were regarded as excellent results.
The longest distance measured between ping-pong balls at the Landiswil test site was
approximately 3Æ9 m which was approximately 70% of the longest distance (5Æ7 m) present in
the volume surveyed. In order to test for extrapolation effects in cases where the object of
interest extends much further than the largest measured distance and where the measured
distances cover only a small part of the photographed outcrop, the Sigma camera was evaluated
on the Illgraben site as well. In this case five images were acquired, two convergent images
aiming from the far left and far right site of the outcrop in landscape format and three images
parallel to the outcrop in portrait format (Fig. 4). The volume covered was much larger than for
the Landiswil site and therefore the image scale and the expected accuracy in object space were
less. The largest LME value was 0Æ0209 m while rmse was 0Æ0094 m including a calibration of
the interior orientation. The Sigma SD14 demonstrated good geometric stability, even though
the best accuracy in object space was accomplished with the bundle adjustment with
calibration of the interior orientation. The camera was calibrated a week after the images were
taken (Table 1). LME values were scattered around zero (Fig. 6) and did not show any
significant scaling problem as seen in the Landiswil data. While the absolute deviations were

25
Length measurement error (mm)

20

15

10

–5

–10

–15

–20

–25
0 4000 8000 12000 16000 20000 24000
Distance (mm)

Fig. 6. Length measurement errors, Illgraben test site. Two convergent images were taken from far left and far
right of the outcrop in landscape format, three images were taken parallel to the outcrop in portrait format. Results
are shown for the Sigma camera using bundle adjustment with calibration of the interior orientation.

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 17
Rieke-Zapp et al. A photogrammetric surveying method for field applications

larger than for the Landiswil site, all points were well scattered around zero. The increase in
absolute deviations can be explained by the larger dimensions (larger image scale number) for
the Illgraben field site. Object coordinates derived from the images were based on distance
measurements covering only a small part of the outcrop area. This will typically result in a
systematic error with increasing distance away from reference points due to extrapolation of
distance measurements. This effect was not visible when plotting the LME (Fig. 6). Plotting
the LME rather indicated that random errors were larger than systematic errors (Fig. 6). The
major source for random errors in the processing chain was the precision of manual point
measurements in the images, which increases proportionally to image scale. Maximum LME
and rmse fell within the accuracy requirements for an outcrop survey of this size. Image
resolution per pixel was just good enough to recognise pebbles in the outcrop. This indicated at
the same time that this 20 · 18 · 5 m3 outcrop defined the largest area that can be covered with
the five-image set-up and the method when working with the Sigma SD14.
The extra weight that had to be carried to the field for the simplest set-up was a laser
distance meter, three ping-pong balls and some modelling clay, as well as a one-page field
manual (Fig. 7). The weight of this equipment adds up to 0Æ175 kg. Considering that a suitable
camera may already be part of a field scientist’s backpack or that a camera such as the Ricoh
GR Digital may replace an existing geometrically less suitable camera, only minimum weight
was added, but accurate measurements were possible. Adding the weight of the Ricoh GR
Digital which is 0Æ2 kg including batteries and memory card, the whole equipment weighs less
than a pound. Since calibration of the Ricoh camera on the job was successful with the

Fig. 7. Digital camera, laser distance meter, ping-pong balls and modelling clay were required for the method in
the field. The red goggles and reference cross were optional.

18  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

five-image set-up, the camera may be used to take overview images with autofocus, and metric
images by adjusting the focus manually to the pre-calibrated setting. Useful extra tools were
red goggles (0Æ05 kg) for better visibility of the laser dot and a weatherproof camera bag. For
cases where the object coordinate system should be aligned with a level plane, a reference
cross with a bubble level was constructed (0Æ2 kg; Fig. 7). Alignment with a level plane can be
important if inclination measurements are needed. Four signalised points were placed on the
cross that can be measured automatically in the images (ellipse operator in Rollei CDW), and
thus allow referencing the X–Y plane instead of using the 3–2–1 method. The distances
between the points on the cross (£0Æ3 m) were considered too short to use them as scale
information. Including all the extras, only around 200 g of not overly expensive field
equipment need to be added for accurate field surveys.
It took less than 15 min to place three ping-pong balls or wooden spheres at either of the
two outcrops, measure the distance between the spheres, take five images and prepare a little
sketch of the set-up. Post-processing time at the computer from import of images in Rollei
CDW to starting the stereo viewer in LPS was approximately 30 min for a user with little
experience in the software. Similar timing was accomplished by students who were introduced
to the methodology and were handed an instruction manual guiding them through the whole
process from data acquisition to image measurement. The methodology was simple, the time
required for image acquisition and measurement was short and the process attracted the interest
of students who started new projects on their own after the first test project was finished.
Considering that accuracy in object space was well within typical demands of outcrop
mapping, the presented method was both accurate and interesting to potential users. The
maximum volume was limited by the resolution of the cameras.

Conclusion
A photogrammetric method was developed aimed at earth scientists wishing to take
accurate measurements in the field while minimising the extra bulk and weight of the
equipment required. In the best case, the equipment will add only 0Æ2 kg to the backpack of a
field scientist. At the first test site in Landiswil a maximum LME between 0Æ005 and 0Æ002 m
was accomplished with all three cameras. For the second test, at a much larger outcrop in the
Illgraben area, the maximum LME increased to 0Æ021 m with an rmse 0Æ010 m. Both results
were considered excellent and were deemed sufficient for a field survey method. The Ricoh GR
Digital was by far the smallest and lightest camera and yielded the best accuracy in object
space when calibrated on the job. The Sigma SD14 was the camera with the best geometric
stability and did not require on-the-job calibration for accurate results. The Rollei d7 metric5
also yielded good results, but could not deliver the extra value that would be expected from a
metric camera designed for photogrammetric surveys. The whole methodology including the
software workflow for image measurement was easily accepted by students and could become
a popular surveying method.
In practice, resolution of fine detail on the outcrops was the limiting factor defining the
largest area that can be covered by the method with a given camera. Although resolution of fine
detail is not solely related to the number of megapixels of a digital camera, the increasing pixel
count of small and lightweight cameras with fixed focal length lenses should allow coverage of
even larger outcrop areas with sufficient accuracy in the future.
The methodology is easily scaled to cover larger areas with more images or may be
refined for applications with higher accuracy demands, such as camera calibration or for
industrial measurements. Potential users may therefore grow with the task at hand and will not
have to learn a new methodology when their project requirements change.

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 19
Rieke-Zapp et al. A photogrammetric surveying method for field applications

references
Fryer, J. G., 1996. Camera calibration. Close Range Photogrammetry and Machine Vision (Ed. K. B. Atkinson).
Whittles, Caithness. 371 pages: 156–179.
ITTF (International Table Tennis Federation), 2007. ITTF Handbook 2007/2008. http://www.ittf.com/
ittf_handbook/ittf_hb.html [Accessed: 7th January 2008].
Lascelles, B., Favis-Mortlock, D., Parsons, T. and Boardman, J., 2002. Automated digital photogram-
metry: a valuable tool for small-scale geomorphological research for the non-photogrammetrist? Transactions
in GIS, 6(1): 5–15.
Leica, 2006. Leica Disto A3 User Manual. Leica Geosystems AG, Heerbrugg. 6 pages.
Leica Geosystems, 2005. ERDAS Field Guide. Leica Geosystems, Atlanta. 770 pages.
Luhmann, T., Robson, S., Kyle, S. and Harley, I., 2007. Close Range Photogrammetry: Principles, Tech-
niques and Applications. Whittles, Caithness and Wiley, New York. 528 pages.
Meydenbauer, A., 1912. Handbuch der Messbildkunst. Wilhelm Knapp, Halle. 245 pages.
Mikhail, E. M., Bethel, J. S. and McGlone, J. C., 2001. Introduction to Modern Photogrammetry. Wiley,
New York and Chichester. 479 pages.
Pike, R. J., 2000. Geomorphometry: diversity in quantitative surface analysis. Progress in Physical Geography,
24(1): 1–20.
Rieke-Zapp, D. H. and Nearing, M. A., 2005. Digital close range photogrammetry for measurement of soil
erosion. Photogrammetric Record, 20(109): 69–87.
Rieke-Zapp, D., Oldani, A. and Peipe, J., 2005. Eine neue, hochauflösende Mittelformatkamera für die digitale
Nahbereichsphotogrammetrie. Publikationen der Deutschen Gesellschaft für Photogrammetrie, Band 14 (Ed.
E. Seyfert). Weinert, Berlin. 482 pages: 263–270.
Rollei, 2002. Close Range Digital Workstation User’s Manual Version 2.1. Rollei, Braunschweig. 235 pages.
VDI/VDE, 2002. VDI/VDE Richtlinie 2634 Part 1. Optical 3D Measuring Systems, Imaging Systems with Point-
by-Point Probing. Beuth, Berlin. 10 pages.
Wackrow, R., Chandler, J. H. and Bryan, P., 2007. Geometric consistency and stability of consumer-grade
digital cameras for accurate spatial measurement. Photogrammetric Record, 22(118): 121–134.
Welch, R. and Jordan, T. R., 1983. Analytical non-metric close-range photogrammetry for monitoring stream
channel erosion. Photogrammetric Engineering & Remote Sensing, 49(3): 367–374.
Wester-Ebbinghaus, W., 1983. Einzelstandpunkt Selbstkalibrierung—Ein Beitrag zur Feldkalibrierung von
Aufnahmekammern. Deutsche Geodätische Kommission, Reihe C, 289. 48 pages.
Wester-Ebbinghaus, W., 1985. Verfahren zur Feldkalibrierung von photogrammetrischen Aufnahmesystemen im
Nahbereich. Kamerakalibrierung in der photogrammetrischen Praxis (Eds. G. Kupfer and W. Wester-
Ebbinghaus). Deutsche Geodätische Kommission, Reihe B, 275, München. 181 pages.

Résumé
Une méthode photogrammétrique est présentée pour la cartographie des
affleurements rocheux sur le terrain. On a veillé tout particulièrement à simplifier la
procédure et à minimiser l’encombrement et le poids du matériel requis tout en
optimisant la précision du levé. L’équipement minimum nécessaire pour les levés a
alourdi de moins de 0,5 kg le bagage d’un scientifique de terrain, ou d’à peine 0,2 kg
dans le cas où l’équipement de terrain comprend déjà un bon appareil photograph-
ique numérique. Il a fallu à un utilisateur peu entraı̂né moins de 15 minutes pour
acquérir les données, et moins de 30 minutes pour orienter les images et préparer
l’observation stéréoscopique. Les meilleurs résultats ont été obtenus en prenant deux
images convergentes ainsi que trois images parallèles à l’objet étudié. La méthode a
été appliquée à deux sites test couvrant un volume de 3,5 · 5 · 3 m3 et 18 · 20 · 5 m3
afin d’identifier la méthode d’ajustement la plus précise. L’erreur maximale de
mesure des longueurs a été calculée pour 28 et 78 distances, respectivement dans le
plus petit et le plus grand volume, par comparaison du levé photogrammétrique avec
un levé indépendant par station totale sur les points signalisés. L’erreur maximale de
mesure des longueurs varie de 0,005 m à 0,002 m sur le premier site test où trois
caméras ont été testées, et vaut 0,021 m pour le second site test où une seule caméra

20  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.
The Photogrammetric Record

a été testée. L’erreur quadratique moyenne des longueurs mesurées varie de 0,001
à 0,003 m pour le premier site et vaut 0,010 m pour le second. La caméra la plus
petite et la plus légère, une Ricoh GR Digital, a fourni les résultats les plus précis
dans l’espace objet avec une orientation interne effectuée sur place. La Sigma SD14
n’a pas nécessité d’étalonnage sur place pour fournir des résultats précis et s’est
avérée être la caméra de meilleure stabilité géométrique. La troisième caméra testée,
une Rollei d7 metric5, a également fourni de bons résultats, sans pour autant
apporter la valeur ajoutée que l’on pourrait attendre d’une caméra métrique conçue
pour les levés photogrammétriques.

Zusammenfassung
Es wird eine photogrammetrische Methode zur Vermessung von geologischen
Aufschlüssen und anderen Objekten im Gelände vorgestellt. Besonderes Augenmerk
wurde darauf verwendet, einen möglichst einfachen Arbeitsablauf zu gewährleisten,
um mit minimaler Ausrüstung ein Maximum an Genauigkeit zu erreichen. Die
photogrammetrische Mindestausrüstung fügte 0Æ5 kg Gewicht zur Geländeausrüstung
hinzu. Geht man davon aus, dass eine Kamera zur Standardausrüstung im
Feldgepäck eines Geowissenschaftlers gehört, und diese auch für photogrammet-
rische Messungen eingesetzt werden kann, beträgt das Extragewicht nur 0Æ2 kg. Die
Datenaufnahme im Gelände betrug 15 Minuten, die Orientierung der Bilder und die
Vorbereitung für eine Stereoauswertung dauerte weniger als 30 Minuten für
Anwender mit wenig Erfahrung. Die besten Ergebnisse wurden erreicht, wenn das
Untersuchungsobjekt mit zwei konvergenten und drei parallelen Bildern erfasst
wurde. Zwei Testobjekte mit einem Volumen von 3Æ5 · 5 · 3 m3, beziehungsweise
18 · 20 · 5 m3, wurden mit der Methode vermessen, um zu ermitteln, auf welche Art
die beste Genauigkeit im Objektraum erreicht werden kann. Die grösste Längen-
messabweichung (LMA) wurde für 28 Strecken im kleineren, beziehungsweise 78
Strecken im grösseren Messvolumen, ermittelt. Die LMA basiert auf einem Vergleich
von Messungen signalisierter Punkte am Objekt aus den Bildern mit einer
unabhängigen Messung der gleichen Punkte mit einer Totalstation. Die grössten
LMA variierten zwischen 0Æ005 m und 0Æ002 m für den kleineren Ausschluss, an dem
Ergebnisse von drei Kameras verglichen wurden, und 0Æ021 m am grösseren
Aufschluss, der nur mit einer Kamera aufgenommen wurde. Die RMSE Werte aller
LMA variierten zwischen 0Æ003 m und 0Æ001 m für den kleineren und betrugen
0Æ010 m für den grösseren Aufschluss. Die kleinste und leichteste Kamera im
Vergleich, eine Ricoh GR Digital, erzielte die beste Genauigkeit am Objekt, wenn die
innere Orientierung simultan kalibriert wurde. Die Sigma SD14 war die geometrisch
stabilste Kamera im Test und benötige keine Simultankalibrierung. Die dritte
Kamera, die getestet wurde, war eine Rollei d7 metric5, welche auch gute LMA Werte
erreichte, jedoch nicht den Mehrwert erbringen konnte, den man von einer
metrischen Kamera erwarten würde.

Resumen
Se describe un método fotogramétrico para cartografiar salientes rocosos y
otros objetos en el campo con el objetivo de simplificar el flujo de trabajo y reducir la
cantidad y peso de los equipos de medida sin por ello comprometer la exactitud del
trabajo de campo. El equipo mı́nimo necesario para los levantamientos supuso

 2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd. 21
Rieke-Zapp et al. A photogrammetric surveying method for field applications

añadir menos de medio kilo al peso de la mochila del investigador de campo, o


solamente 0,2 kg suponiendo que un equipo tı́pico para observación en campo ya
incluye una cámara. A un usuario poco entrenado, la obtención de datos le toma
menos de 15 minutos y la orientación y preparación de la imagen para la visión
estéreo menos de 30 minutos. Los mejores resultados se obtuvieron tomando dos
imágenes convergentes ası́ como tres imágenes paralelas al objeto de interés. Para
determinar el método de ajuste más exacto, se realizaron ensayos en dos zonas de
3,5 · 5 · 3 m3 y 18 · 20 · 5 m3. A continuación se calculó el error máximo de medida
de la longitud (EML) en 28 y 78 distancias en los volúmenes más pequeño y más
grande respectivamente, comparando el levantamiento fotogramétrico con un
levantamiento independiente realizado con estación total de los mismos puntos
señalizados. El EML osciló entre 0,005 y 0,002 m para la primera zona de ensayo en
donde se probaron tres cámaras, y 0,021 m en la zona más grande donde sólo se
probó una cámara. Los valores rmse de los EML oscilaron entre 0,001 m y 0,003 m
en la primera zona, y 0,010 m en la segunda. La cámara más pequeña y ligera, una
cámara Ricoh GR Digital, proporcionó los resultados más exactos en el espacio
objeto cuando se calibró la orientación interna sobre el terreno. La cámara Sigma
SD14 no requirió de calibración sobre el terreno para obtener resultados exactos y
fue la cámara con la mejor estabilidad geométrica. La tercera cámara evaluada, una
Rollei d7 metric5, también proporcionó buenos resultados, pero no aportó la
capacidad extra esperable en una cámara métrica diseñada para trabajos
fotogramétricos.

22  2009 The Authors. Journal Compilation  2009 The Remote Sensing and Photogrammetry Society and Blackwell Publishing Ltd.

You might also like