HIGH PRECISION CALIBRATION OF CLOSE RANGE PHOTOGRAMMETRY SYSTEMS 
Hädem, l., University of Trondheim, Trondheim, Norway 
Amdal, K., Metronor A S, Oslo, Norway 
Commission V 
ABSTRACT 
This paper deals with investigations on test field calibration of close range photogrammetric systems for high precision industrial 
applications, with emphasize on the use of stereo vision systems of 2 or 3 cameras. The initial investigations are based on 
simultation studies: The influence of different parameters on the accuracy of the calibration are investigated i.e. the number and 
configuration of signalized object points, the object control of given points or given distances, the number and configuration of 
exposure stations, and the type of camera parameters. Finally, the result of a practical investigation on the calibration of a digital 
photogrammetric system on the basis of given distances in object space is reported. 
KEY WORDS: Accuracy. Calibration. Close-range. Photogrammetry. 
1. INTRODUCTION 
Traditionally, the objective of calibration is to estimate those 
parameters in the photogrammetric system which can be 
considered as "constants" in later photogrammetric 
measurement tasks. The parameters of the photogrammetric 
system (the functional relationhip between image points 
(x,y) and object points (X,Y,Z)) are primarily those of inner 
and outer orientation describing the fundamental model: the 
central perspective, and secondly additional parameters which 
describe the deviations from this model. These deviations 
(model errors) can conveniently be mathematically 
formulated as systematic image errors, on the basis of a 
physical approach or a numerical/statistical approach. The 
systematic image errors are often considered as belonging to 
the concept of "inner orientation" in a wider definition of this 
concept. It should be mentioned that when using analog 
cameras the inner orientation has to be restored for every 
picture, on the basis of fiducial marks. Such a restoration is 
not necessary when using digital cameras, as it can be assumed 
that the position of the pixels in the image system retain their 
positions from exposure to exposure. In some cases the 
calibration may also include the outer orientation, as for 
instance when a permanently mounted unit of cameras is used 
for a consecutive dimensional control of constructions in a 
workshop hall. Considering a stereo vision system as a unit of 
cameras which have a fixed relative orientation (relative 
camera rotations and positions), this orientation is also 
subject to calibration. 
The simulation approach has generally become very popular to 
guide the surveyer in network design. Within close-range 
photogrammetry simulation studies of the factors influencing 
the accuracy and reliability are reported, together with 
results of verifying the conclusions on applications. (Fraser, 
1989 and Schlógenhofer, 1989). 
The primary objective of this paper is on the basis of 
simulation to investigate some factors which influence the 
accuracy of test field calibration. Such factors may be the 
configuration of object points and camera orientations, inner 
orientation including systematic errors (local or global), and 
the object control (given points and distances). As a measure 
on the accuracy of the calibration the relative accuracy of 
estimated unknown distances in representative positions and 
directions within the test field is given. (The variance/ 
covariances of the etimated calibration parameters might also 
have been used). The simulated cases are rather restricted; 
further investigations are therefor highly desirable. At the 
end real results of calibrating a high precision digital 
photogrammetric system are reported. However, first some 
aspects of precalibration are discussed. 
2. PRECALIBRATION METHODS 
There are two main approaches to precalibration: Optical 
calibration and Test field calibration. (Freyer, 1989). As the 
simulation in this paper deals mainly with test field 
calibration, the optical calibration is briefly discussed. 
2.1 Optical calibration 
Optical calibration uses optical means for a thorough 
laboratory test of the physical and mechanical function of the 
camera and its geometrical quality, including a high precision 
estimation of the inner orientation parameters, radial lens 
distortion (often in the 4 diagonal directions) and sometimes 
also decentering. (See e.g. Burner, 1990). For analog 
cameras, the calibrated coordinates of fiducial marks serve as 
a means for restoring the calibration image system (where 
the principle point is defined) in later photogrammetric 
measurment tasks. In principle a transformation on the 
fiducial marks by translation and rotatation should suffice. A 
more sophisticated transfomation may, however, model 
physical sources of errors (like film shrinkage) which have 
been active between the exposure and the measurement of the 
image. 
2.2 Test field calibration 
The actual calibration parameters are estimated on the basis 
of measuring pictures taken of a test field. The dependence of 
principle distance, radial distortion and decentering on the 
focussing (which in turn is dependent on the distance between 
the object and the exposure station) may also be subject to 
calibration, (Freyer, 1989). The calibration conditions 
(temperature, illumination, .. ) should be as possible similar 
to those expected in later photogrammetric tasks. The 
disadvantage of test field calibration is the large quantity of 
work involved: A stable 3-dimemsional steel frame with e.g. 
retro-reflex targets must be built, and next a control must be 
established. The control may be given points (X,Y,Z) in some 
chosen object system and/or given distances between some of 
the targeted points. Given distances may also be introduced by 
placing bars with targeted points (which are the endpoints of 
the given distances) in favourable positions and directions 
within the actual object space. The control may be established 
by high precision geodetic methods. The control of a test field 
for calibrating digital cameras of moderate accuracy may, 
however, be photogrammetrically established using high 
precision metric analog cameras (Amdal et al, 1990). 
Equivalent to the use of a 3-dimensional frame is the use of a 
2-dimensional frame which can be positioned paralell to itself 
(Beyer, 1987).