— The relative image resolution is lower, which results
in a smaller object area covering and therefore
leads to a greater number of cameras. The effort to
be made for orientation and calibration increases.
— In most cases every image is taken with a unique
camera. This reduces the field calibration capability.
Also every additional image is quite expensive.
Therefore redundancy is often poor.
— Orientation elements for online system checks must
be provided.
— Limitations in the positioning of cameras and con-
troller hardware are given by cabling.
A good optimization has to be done in order to come to
efficient systems.
1.4 The problem of system evaluation
For the quality control manager who has to decide if a
photogrammetric video system fits to his requirements
arises the problem of correctly evaluating the available
solutions. A formalism with a set of comparable parame-
ters does not exist. The dependence on individual con-
figuration aspects is very high. Some suggestions will be
made at the end of this paper.
2. CONFIGURATION AND ORIENTATION
Determinating the orientation of an image is a basic pho-
togrammetric operation and necessary for most of the
evaluation methods. It can be done online with additional
object information, offline by calibration or indirectly with
special hardware. It influences configuration considera-
tions heavily.
2.1 Online-orientation versus calibration
In a static measurement system it is possible to calibrate
fixed camera positions in a seperate step instead of ori-
enting every new image. It depends on the mechanical
and electronical stability of cameras and fittings relative to
environment conditions, how long uncontrolled operation
is guaranteed to be errorfree. The advantages are:
no additional targets neccessary
better independence from object variation
higher accuracy of orientation
faster evaluation
The combination of both orientation methods where the
separate calibration is periodically controlled online pos-
sibly with a reduced set of parameters is of course the
best solution.
2.2 Aspects of accuracy, speed and economy
The accuracy of measurement results influences the
configuration through the image scale, which together
with the camera resolution defines the object area cov-
482
ered by one image. It is important to clarify very precisely
which accuracy is really needed, because every addi-
tional image may become expensive, because in static
online measurement systems with fixed camera positions
this will require a new camera and possibly new or ex-
tended controller hardware. For this reason it depends
much more on the block configuration whether a system
is economical or not than in conventional photogramme-
try. Another factor for costs is the speed of image acqui-
sition and evaluation at the point where the job cannot be
done in time by standard hardware. Concerning the cam-
era configuration it may limit the use of moving compo-
nents which are discussed more detailed below.
2.3 Flexibility
i
In most cases a static video measurement system is
designed for exactly one well defined task. Nevertheless
the requirements given by the industrial production proc-
ess may change or the system has to be placed to an-
other location or it has to be adapted to different objects.
In these cases a very specialized and optimized configu-
ration shows low flexibility. This has to be taken into ac-
count when reducing camera numbers and using uncon-
ventional orientation methods which are described below.
2.4 Configuration types
A video measurment system may have the following prin-
cipal camera constellations:
a) 1 camera
b) 2 cameras
C) several cameras in block configuration
d) several independently oriented cameras
a) One single camera can be used to determine object
coordinates in a plane.
b) Using two cameras with overlapping image areas is
the simplest way to obtain three-dimensional object coor-
dinates. There is no redundancy for error control and the
object size is limited if high accuracy is needed.
c) The block configuration, that means several overlap-
ping images which cover the whole object, is the most
accurate constellation, supposed that a bundle block
adjustment with sufficient redundancy is used. It is very
flexible, but it needs a great number of tie points which in
most cases have to be signalized on the object surface.
The number of required cameras is high. If an object has
only a few points of interest with long distances between
them it may occur that several images only act as block
connections, which is very ineffectiv and expensive.
d) Independently oriented cameras can concentrate on
the important parts of the object. It has to be considered
that every target must be measurable in at least two im-
ages, which leads to a pair or better a triple of cameras
for each interesting object region. This method is dis-
cussed in detail below.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
2.5 Cc
A simj
enoug|
resecti
oriente
cal poi
This o
justme
placed
have tc
positio
by the
camer:
the obj
does n
approa
2.7 Sp
The lin
produc
positio!
small «
tances
unfavol
accura
and co
sibly cz
manen
Operati
optimiz
mum o
mainly
3.1 Obj
As mer
the whc
nation :
which a
comple
figuratic
when a
ready p
task to
really ir
eliminat
remainii
This ca
parts wi
trol poir
bilities ;
for this
dance e
Also the
images
ages he
conditio
