MONTE-CARLO-SIMULATION IN CLOSE-RANGE PHOTOGRAMMETRY
Heidi Hastedt
Institute of Applied Photogrammetry and Geoinformatics, Oldenburg, Germany, h.hastedt@vermes.fh-oldenburg.de
Commission V, WG V/1
KEY WORDS: Accuracy, Adjustment, Calibration, Metrology, Modelling, Photogrammetry, Simulation
ABSTRACT:
The process chain in optical measurement techniques can be subdivided into four main components: the camera system, the object
range, the network design and the analysis system. The included influences (e.g. camera geometry, illumination, algorithms for
image measurement) cause remaining deviations on the results due to insufficiently known effects on the photogrammetric system.
This article will introduce a simulation technique based on Monte-Carlo-Methods to analyse effects of camera geometry, object
space, signalisation and illumination. First two topics will be discussed based on simulation results. It allows a closer look at single
system components, their uncertainty and randomly distribution simultaneously to the estimation of their influence on the
photogrammetric system. The described Monte-Carlo-Simulation provides an economical process where the effects can be separated
and modelled within an acceptable period of time and amount of work. It enables the determination of optimal system components
(e.g. signalisation, illumination, camera geometry, analysis) and, in addition, the estimation of their influences on the process chain
due to given (fixed) system components.
1. INTRODUCTION
The process chain in optical measurement techniques can be
subdivided into four main components: the camera system
(camera geometry, illumination, the object range
(configuration, complexity, signalisation), the network design
(configuration, scales, control elements) and the analysis
system (algorithms for image measurement, functional model
for camera geometry and bundle adjustment). Caused by this
complex process chain photogrammetric results include
remaining deviations due to insufficiently known effects.
Nowadays used digital high-resolution consumer cameras do
not remain stable within an acceptable period of time, not
within the period of image acquisition either. Therefore a new
camera model was discussed and verified by Hastedt et al.
(2002). An image-variant interior orientation is added to the
functional model, which describes variation in principal
distance and principal point. In order to compensate sensor
based influences and remaining lens effects not considered
within radial-symmetric lens distortion, a finite-elements sensor
correction grid has been chosen. The mentioned camera model
enables the use of instable digital high-resolution cameras for
high precision purposes.
Choosing the right object range for calibration and verification
purposes, the German Guideline VDI/VDE 2634,
recommending a special configuration, gives particular support.
Rautenberg & Wiggenhagen (2002) discussed the verification
of different optical measuring systems based on this guideline.
Hastedt et al. (2002) followed up this verification concept and
demonstrated remaining length dependent deviations within the
length measuring error.
In case of industrial measurement techniques retro-reflective
material is used for signalisation combined with the use of ring-
lights. Dold (1997) demonstrated the problem of this material.
In particular the marginal reflection is affected and does not
meet the required exact reflection. The choice of the material is
an important component of the photogrammetric process,
particularly regarding the subsequent measuring algorithm.
The optimization and specification of the network design has
been discussed in several publications, e.g. Fraser (1984),
Zinndorf (1986). Fraser (1984) explained the dependence on the
Datum Problem (Zero-Order Design) and the Configuration
Problem (First-Order Design). Regarding the optimization of
the network design previous investigations and applied
approaches have to be modified for recently used methods and
new digital equipment and its advantages of flexible system
components. One step towards this modification constitutes a
simulation tool designed for special applications in crash-
techniques, which has been developed by Raguse &
Wiggenhagen (2003).
Having a closer look at the analysis system, two components
are mainly influencing the systems result. First, belike one of
the most important system parts, the algorithm (template
matching, ellipse operator) measuring the centre of the imaged
point mark has to be addressed. The importance of its influence
is insufficiently known. Secondly the earlier described camera
model.
In order to gain the single forces of the described components
in an economical process where the effects can be separated
and modelled within an acceptable period of time and amount
of work, a simulation technique based on Monte-Carlo-Methods
has been developed and will be introduced by this article. The
simulation method allows a closer look at single system
components, their uncertainty and randomly distribution
simultaneously to the estimation of their influence on the
photogrammetric system. The analysis of the simulation results
of this report will focus on the influence of the camera
parameters and geometry as well as on the influence of the
object space, herein the systems exterior.
Internat
The Me
techniqi
of non-i
chosen.
with the
simulati
systems
problem
randomi
getting
(Schwei
evaluati
includes
density
g(X)=
X =(X
The pdf
inputs t:
Simulati
chart foi
js
Figure
After sj
appropri
selected.
univaria
has bee
generato
are the
(Schmitt
generati
like Hil
Muller /
the stanc
IU, U
distribut:
defined |
xX, = 4-
X= of
