ons 
n a 
be 
vith 
1 of 
ble 
FG 
and 
les. 
] to 
ctly 
the 
and 
cell 
l. It 
Ses 
the 
  
89 
These cell parameters are defined by photogrammetric bundle adjustment. With a still video camera (KODAK DCS420) a 
photogrammetric network of 39 images is taken. The réseau crosses are imaged through the upper opening of the cell to 
avoid a multi-media path through the windows. To connect these images with the images made from the front side of the 
fluid cell and to get a stable geometry for the network a field of tie points is fixed on the whole fluid cell. 
The images are measured with the digital photogrammetric station DPA [Bósemann et al. 1994]. Due to automatic 
measurement operators the image coordinates of the réseau crosses and the control points are measured with subpixel 
accuracy. The accuracy of the bundle adjustment defined by oy is about 0.5um. The standard deviations of the object 
coordinates of the réseau crosses and the control points for the camera orientation are in the range of 6um to 11um. 
The scale of the photogrammetric evaluation of the fluid cell is defined by a pre-calibration of the réseau crosses. The 
coordinates of the réseau crosses were automatically measured with a monocomparator with an average accuracy of 
about 3 um. These coordinates are used in the bundle adjustment as additional observations. 
The three-dimensional coordinates of the réseau crosses of both réseau planes and the control points on the front side of 
the cell are now available in the previously defined coordinate system. 
3.2 Calibration of the photogrammetric front end 
Photogrammetric evaluations with an object accuracy of some micrometers require highest accuracy when describing the 
projection process. Deviations from the ideal central projection have to be modeled by error describing functions and have 
to be taken into consideration when determining unknown parameters. Therefore the imaging system has to be calibrated 
before evaluating any fluid experiment. The focal length, the principal point and parameters to describe lens distortion and 
affinities caused by camera synchronisation have to be determined. The calibration process is based on the 
photogrammetric bundle adjustment. Some additional error describing functions are added to determine the interior 
orientation parameters of the cameras with the exterior orientations. 
A testfield with about 30 retro reflective control points fixed on two planes is imaged by all cameras of the camera head 
from about eight different positions [Godding 1993]. These eight positions are simulated by rotating and turning the 
testfield instead of changing the camera position. Because of the fixed relative orientations of the cameras the 
measurement of the image coordinates can be fully automated. The unknown parameters are determined automatically in 
an iterative process of bundle adjustment and by removing blunders from the image observations. 
The result of the calibration process are the parameters which describe the complete orientation of the cameras in the 
testfield coordinate system. 
4. EVALUATION CONCEPT 
The determination of the position of the liquid drops and columns and their shape are achieved in two steps: first the 
calculation of the projection rays through the different media, and second, the determination of the bodies in the liquid. 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop “From Pixels to Sequences", Zurich, March 22-24 1995