Full text: XVIIIth Congress (Part B5)

/ery precisely 
' every addi- 
use in static 
lera positions 
y new or ex- 
n it depends 
her a system 
hotogramme- 
image acqui- 
ob cannot be 
ing the cam- 
ving compo- 
OW. 
nt system is 
Nevertheless 
duction proc- 
laced to an- 
arent objects. 
ized configu- 
aken into ac- 
using uncon- 
ribed below. 
ollowing prin- 
1 
as 
ermine object 
age areas is 
| object coor- 
ntrol and the 
sd. 
'eral overlap- 
is the most 
bundle block 
ed. It is very 
ints which in 
ject surface. 
in object has 
ces between 
act as block 
ansive. 
ncentrate on 
e considered 
least two im- 
> of cameras 
ethod is dis- 
2.5 Conventional orientation methods 
A simple way of determining an orientation is to provide 
enough control points in the image and calculate a spatial 
resection. In a block arrangement the images can be 
oriented relatively, whereby it is necessary to have identi- 
cal points which do not need to have known coordinates. 
This operation can be performed by a bundle block ad- 
justment. In both cases additional targets have to be 
placed in object space, in the case of control points they 
have to be determined in a separate calibration step. The 
positions of such targets of course must also be covered 
by the images, which may sometimes lead to additional 
cameras. If point signalization on the surface or in front of 
the object is not possible and an offline system calibration 
does not come into consideration, a different orientation 
approach has to be done. 
2.7 Special demands in industrial environments 
The limit for configuration layouts given by the industrial 
production process is mainly that the optimal camera 
positions are not accessible. The consequences can be 
small object regions covered by images for short dis- 
tances between camera and object and in most cases 
unfavourable ray intersections which lead to reduced 
accuracy. Important for the stability of camera positions 
and control points are production activities or traffic, pos- 
sibly causing that orientations have to be controlled per- 
manently. 
3. OPERATING RANGE ENLARGEMENT 
Operating range enlargement in this context means to 
optimize extensive configurations with the goal of a mini- 
mum of cameras and a maximum of object size. It is 
mainly a problem of camera orientation. 
3.1 Object coverage 
As mentioned before it is not always necessary to cover 
the whole object with images. An example is the determi- 
nation of profiles of a railway waggon of 18 m length, 
which are only required in the middle and at both ends. A 
complete coverage with a usual close range block con- 
figuration would only be possible with a lot of cameras, 
when an accuracy of 1 mm has to be achieved. As al- 
ready pointed out many of these cameras only have the 
task to build the block structure and do not show any 
really interresting object point. When such cameras are 
eliminated, the block becomes instable or unusable. The 
remaining cameras have to be oriented independently. 
This can be done in the same coordinate system for all 
parts which is represented for example by a field of con- 
trol points determined in a special calibration. The possi- 
bilities and advantages of block adjustment are reduced 
for this type of configuration because of lower redun- 
dance and slighter connections between the block parts. 
Also the elimination of tie cameras can result in remaining 
images that have no orientation information. These im- 
ages have to be oriented by a calibration, or in unstable 
conditions with alternative methods described below. 
483 
3.2 Moving cameras 
Camera movement in an otherwise static measurement 
system can be used to obtain more images without addi- 
tional video hardware. This is the same idea as the macro 
scanning used in some digital photogrammetric cameras. 
There are two principal movement types which have dif- 
ferent effects: the rotation and the shifting. 
Rotating cameras simulate a greater field angle. They can 
cover a larger object region. Depending on layout and 
precision of the fitting the images taken in different direc- 
tions have none or up to five identical orientation pa- 
rameters. The rotation parameters are given by the the 
movement itself, whereby the cameras normally do not 
roll around their optical axis. Position shifts are caused by 
different locations of the rotation and the projection cen- 
ter. This is a mechanical problem, but the effects can be 
calibrated. Two different shots taken from an only rotating 
camera of course cannot be used alone for a three- 
dimensional evaluation because of the missing image 
base. 
Shifted cameras deliver fully independent images. Orien- 
tation parameters may still be partly identical, suitable 
mechanics presupposed. 
A combined movement system with rotations and shifts 
can be very flexible, but loses the advantage of identical 
orientation parameters. 
The disadvantages of moving cameras are the additional 
time consumption, the sensitivity for disturbance and the 
mechanical wear. 
4. ALTERNATIVE ORIENTATION METHODS 
Standard methods use their own image information for 
the orientation process. The disadvantage is that addi- 
tional targets have to be placed in object space, which 
can be difficult, instable or simply impossible. In these 
cases the orientation of cameras can be performed by 
alternative methods which will be described in detail be- 
low. 
4.1 Laser 
To obtain all orientation parameters with a laser system is 
possible but very complicated and inefficient. The ad- 
vantages lie in rotation control and determination. A laser 
attached to the camera body can point to a far target 
outside of the object space, where displacements of the 
beam are equivalent to certain rotations of the camera. 
The check can be performed manually, for example with a 
scale, or an additional camera can be used for the auto- 
mation of this step. Another possibility to control the beam 
movement is given by PSD's (PSD - Position Sensitive 
Device). These elements have a size of 10-30 mm? and 
can register the position of a light point in one ore two 
dimensions with high accuracy. 
4.2 Orientation cameras 
The fact that two cameras fixed together have sets of 
orientation | parameters, which are directly dependent 
from each other, can be used to place control points far 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.