Using the IAPG software package FiBun the above mentioned 
image block could be oriented successfully. As known from 
other digital cameras such as Kodak DCS 460 a significant 
displacement of the principal point could be determined and 
modelled (Fig.3). The adjusted sensor correction grid is 
displayed in Fig. 4. The large deformation are probably caused 
by a heating effect of the internal MicroDrive hard disk (see 
Ch. 4.1). The final results of testfield calibration are listed in 
Table 1. 
  
Sensorplot DCS Pro Backg4645M mit 35 mm Objektiv 
Korrekturgitter 2 Fi krometer a priori 
Fig. 4: Resulting correction grid after bundle adjustment 
The results show quite satisfying results. Since no camera- 
specific handling has been applied during image acquisition, 
and no special adaptation of target size or illumination has been 
carried out, the overall accuracy of 0.01 mm in object space and 
0.33um in image space demonstrate the high potential of the 
camera. The a priori expected level of accuracy has been 
achieved without any further effort. 
The reported inner accuracy corresponds to a relative precision 
of 1:200000. In comparison to a Kodak DCS 460 camera (3000 
x 2000 pixel) or the Fuji FinePix S1 (2304 x 1536) the Kodak 
Pro Back 645M shows a gain of accuracy in the order of 30% 
which is equivalent to the increased image resolution. 
3.3 Error of length measurement 
The validation of the exterior accuracy of such a camera is a 
non-trivial task. Since there are only few measuring techniques 
that provide equal or even higher accuracies (e.g. 
interferometric laser tracking) the exterior accuracy can only be 
verified efficiently by independent, calibrated scale bars. 
According to VDI 2634 there are investigations on how the 
recommended procedures of acceptance and reverification can 
be handled in practice. 
Recent investigations show that the accuracy of a 
photogrammetric length measurement is degraded with respect 
to the theoretically expected value. Rautenberg & Wiggenhagen 
(2002) report on a factor of 2 to 3 for the lack of accuracy if 
calibrated high-precision scale bars are measured by 
photogrammetry. Although this investigation summarises a 
number of practical tests it is still influenced by a number of 
unknown parameters such as the effect of retro-reflective 
targets, or inhomogeneous accuracy behaviour in object space. 
The first tests using the Pro Back system show the same effect 
in distance measurement as already known for other digital 
cameras. As a result, an absolute error of length measurement 
of 0.1mm has been evaluated. A more detailed report on this 
approach is given by Hastedt et al. (2002). 
4. PHOTOGRAPHIC CHARACTERISTICS 
The following investigations deal with radiometric properties of 
the CCD sensor and camera-integrated firmware for image 
processing. Recent test results are presented that give a first 
impression of the image quality and some of the effects that 
influence the final result. 
4.1 Dark current noise 
Dark current noise describes the basic electronic noise in terms 
of greylevel deviation that occurs in total darkness. Although 
this does not reflect a typical imaging condition it shows the 
quantitative effect of sensor noise. In cases of very long 
exposure times, and for dark objects (like black target 
background) CCD sensors produce radiometric artefacts like 
so-called hot pixels in addition to image noise and dark current 
noise. It is known that these effects are also a function of 
temperature of the CCD and the electronic devices. 
For the following test the camera lens (AF 3.5/50) was covered 
totally by a black opaque piece of cloth. The viewer was also 
cut off from light. A number of images has been taken with f- 
stop 22, switched-off auto-focus and an exposure time series of 
30s down to 1/4000s. The sensor sensitivity was varied between 
125, 200 and 400 ISO (ASA). 
  
  
  
  
Fig. 5: Example of dark current noise (magnified, contrast 
enhanced, negative color table) 
Fig. 5 shows one example of the resulting image. It can be 
shown that longer exposure times and higher sensor sensitivity 
yield to higher noise and numbers of hot pixels, respectively. 
The temperature behaviour can be addressed as one of the most 
significant effects to the radiometric quality. As illustrated in 
Fig. 6 the upper left image corner (lower right sensor corner) is 
subject to a local radiometric change that could be identified as 
the result of the heat-up of the internal MicroDrive hard disk 
which is positioned in the lower right area of the camera back. 
This effect does not only affect the radiometric quality but, it 
also causes unacceptable geometric deviations in image space 
—dàd— 
(cc 
str 
the 
The 
use 
cor 
WO: 
test 
Ex] 
me: 
gen 
pro 
add 
five 
dov 
4.2 
On 
abil 
rec 
bee 
this 
4A) 
Adc 
pho 
blo: 
con 
One 
edg 
Fig. 
The 
ran 
top 
witl 
chai 
(dar 
sho 
600 
ima 
ove 
pert