ed on
Its for
posed
uracy
Tor is
uture.
mage
ectly
VHS
outh.
/ the
lages
y the
309
Fig. 11 and Fig. 12: Corresponding images taken with left and right camera.
S. DSM GENERATION
DSM is the digital tooth model. The purpose of DigiDent is the measurement of three dimensional position and rotation changes
of every single tooth. The generation of a complete DSM is not necessary on principle. Theoretically geometric parameters of
position and rotation changes can be derived by estimating transformation parameters using four or five homologous points on
every tooth at different time periods. Since no points that can be detected as homologous points can be marked on the teeth a
different solution has to be found. Therefore a DSM of every tooth is generated. Transformation parameters are derived by
comparing the DSM at two different time periods.
For DSM generation until now only image data acquired from a dental test model was used, since questions of illumination have
not been solved until now. However, the same configuration of teeth, mirror and cameras was used. Imagery was acquired
sequentially with only one camera and one frame grabber, since only one imaging system had to be calibrated. Figures 13 and 14
show imagery acquired from the model without special illumination. The crosses on the mirror surround the teeth. Since the
crosses have a line width of only 2-3 pixels they cannot be seen very well in the figures shown below.
es
Fig. 13 and Fig. 14: Stereo pair taken from the dental test modell.
5.1 ORIENTATION OF IMAGES
Images are oriented by measuring the crosses in left and right image. They shall be detected automatically (at the moment they are
measured semiautomatically using least squares template matching) and measured by least squares template matching. Using
bundle adjustment unknowns of exterior orientation are estimated. The reliabilty of the elements of the exterior orientation is very
important, since geometrically constrained least squares template matching is applied for DSM generation. Therefore weakly
matched crosses have to be detected automatically. Robust estimation models have to be impiemented when crosses are detected
and measured automatically in the future. :
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop “From Pixels to Sequences’, Zurich, March 22-24 1995