2. CONCEPTUAL ASPECTS 
2.1 Hard- and Software 
The system consists of two PULNiX TM765 E black/white 
CCD cameras that are equipped with 16 mm lenses. The 
cameras are connected to a standard low cost video 
capture board miroVIDEO D1, that allows image 
acquisition from two different signal sources sequentially 
in video rate. Therefore there is no need for an expensive 
frame grabber that allows | image acquisition 
simultaneously. Both cameras are equipped with a 
special illumination device, that consists of a ring of red 
LEDs. It is used to illuminate both the interior of the 
mouth and retrotargets fixed on a special mirror that is 
necessary to make image acquisition possible. For a 
detailed description of the mirror see chapter 4. 
The video capture board is mounted in a standard PC 
that is equipped with a PENTIUM™ 75 MHz CPU. All 
software developed is written in standard C under 
Microsoft Windows™. 
2.2 Theoretical Concept 
Derivation of 3D position and orientation changes of 
every single tooth using a digital photogrammetric close 
range system needs a special technique. Many 
constraints originating in the nature of the problem make 
it very difficult to find an appropriate way to derive 
deformation parameters. 
As described above the measurement system consists of 
two cameras and a video capture board that allows more 
less simultaneous image acquisition. The system has to 
be calibrated to be used for measurement purposes (see 
chapter 3.5). To make image acquisition possible a 
special mirror has been developed that is placed between 
upper and lower jaw. The mirror shows 11 retroreflective 
targets that are used for relative and absolute camera 
orientation. On the other hand no targets can and may be 
fixed on the teeth. To derive 3D deformation parameters 
three well defined points on every tooth are necessary. 
They may not be destroyed during an orthodontic 
treatment that lasts for several month. Obviously the use 
of artificial target points on the teeth's surface is not 
possible. Therefore deformation parameters are derived 
by 3D DSM-matching. In every time period of the 
orthodontic treatment a complete DSM (Digital Surface 
Model) of every tooth is processed. The deformation 
parameters can be derived by 3D matching of the DSM 
of the same tooth in different time periods. DSMs are 
derived from a stereo pair only by the use of the natural 
structure of the teeth surface. Image acquisition and 
proper illumination in the very close range of the human 
mouth is difficult enough. The use of structured light 
methods might lead to more accurate results but on the 
other hand it might introduce additional problems during 
image acquisition. 
3. SYSTEM ANALYSIS 
Digital close range systems have to be analysed and 
calibrated before they can be used for measurement 
purposes. For the user, who is usually not an expert in 
photogrammetry, it is very important to know how long 
the warm up effect of the system lasts. The 
photogrammtrist is much more interested in the 
248 
performance analysis of the system including noise, 
repeatability and especially system calibration. 
3.1 Repeatability 
For this investigation as well as for the analysis of warm 
up effects images from a small testfield of 150 x 150 mm 
size showing 49 black targets on white background were 
taken. For repeatability analysis a set of five images was 
taken in video rate. The image coordinates were 
measured using LSTM. The averaged coordinate set of 
all five images was used as reference. No systematic 
trend was excluded. The average RMS values for the 
coordinate differences in x and y direction are 0.018 pixel 
and 0.019 pixel respectively. This corresponds to 1/55th 
of a pixel. 
In a second test six images were taken sequentially every 
10 minutes. Again an averaged image coordinate set was 
used as reference but systematic trends were excluded. 
In average the RMS values for the coordinate differences 
in x and y direction are 0.025 pixel and 0.024 pixel 
respectively. Taking into consideration that the frame 
grabber used is a low cost product for approximately 300 
US$, a repeatability of 1/40th of a pixel over on hour is 
pretty good. 
  
  
  
  
  
  
  
  
Version | Trendx | Trendy | RMSx | RMSy 
[Pixel] [Pixel] [Pixel] [Pixel] 
5 frames 0.018 0.019 
1 hour | 0.024 0.018 0.025 0.024 
  
  
Table 1: Results of repeatability analysis 
3.2 Warm Up Effect of Camera 
To achieve best results it is very important to know how 
long the warm up effect of an image acquisition system 
or any other electronic device lasts. The investigation is 
done in two steps: analysis of the camera and analysis of 
the frame grabber. To investigate the camera the frame 
grabber was turned on several hours before the test. 
After the camera was turned on 10 images were taken 
within 2 hours. The last image is used as reference, 
systematic trends are excluded. Figure 2 shows, that the 
warm up effect is not over after 2 hours. The RMS values 
are 1/30th of a pixel compared to 1/40 th of a pixel when 
the system is completely warmed up. 
  
  
    
  
  
0,5 
0,25 
= 0 
® 
X 
a. 
025 —@— trend x [pixel] 
—B-— trend y [pixel] 
08 —A— RMS x [pixel] 
—>— RMS y [pixel] 
0,75 ‘ 
0 1000 2000 3000 4000 5000 
time [seconds] 
Figure 2: Warm up effect of camera 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
3.3 Warm U 
Like every 
warm up ef 
hours. The 
The last im 
shifts are e» 
the frame c 
RMS value 
warmed up : 
0,1 
-0,1 
[pixel] 
-0,2 
  
-0,3 
Fig! 
3.4 Noise A 
To analyse ! 
images was 
was used 
differences 
suprisingly c 
  
Fig 
3.5 System 
The system 
150 x 150 m 
background. 
LSTM. The 
images was 
degrees. E\ 
images. To 
additional ps 
determinable 
estimation | 
coordinate i 
correspondil 
7500)