, Ww wow 
ADM UG UOI AR Te 
"Ww LR 
177 
So, let us shortly recall how accurate and fast the photogrammetric measurement system should be. Basically, there is 
a general trend in industry to speed up the production time of design models. As far as the car industry is concerned, 
this means that for instance a full scale model has to be digitized within one week, preferably 3 days including all 
process steps starting from scratch. For the same application field the accuracy requirements are varying from + 0.1 
mm to £ 2 mm on the full scale model. Actually, it depends on the individual workflow environment and the CAD- 
system(s) in use, which accuracy requirements have to be fullfilled. Also, the accuracy changes during the individual 
design steps of a car model. So far, in most cases the "0.1 mm" accuracy level was requested, even in the sense of a 
tolerance measure, meaning that the related standard deviation has to be in the order of 30 microns. Hence, for such 
an accuracy level a high angular resolution is needed in the digital measurement system, so that only analog cameras 
in combination with a scanner or high resolutional digital cameras (e.g. macroscanning cameras) can be successfully 
applied. 
However, car design strategies are also changing. Modern CAD-systems make it possible to design cars almost only 
in the computer, using powerful visualization techniques like reflections highlighting to get realistic feedback and aes- 
thetic impressions. Thus "physical" design on a full-scale model is partly replaced by "virtual" design in a CAD-system, 
which saves time consuming design steps and speeds up production. In such a flexible CAD environment the impor- 
tance of the initial physical model is reduced, and thus also the accuracy for the digitization changes. The specific 
accuracy requirements of such modern approaches are in the order of + 1 mm on the full scale model. This makes 
low resolutional digital cameras like still video cameras applicable, if the digitization starts for instance on a 25 % car 
concept model, which means that the accuracy has to be in the order of + 0.25 mm at this scale. 
Since their introduction still video cameras like the KODAK DCS 200 were successfully applied in various fields of 
close range photogrammetry (Mass et al., 1994; Peipe et al., 1993) and quickly gained a good reputation as a precise 
measurement tool which is as easy to handle as a simple photo camera. Apparently, they are already in use in car 
industry as visualization and documentation tool. Still video cameras provide on-line digital images which can be im- 
mediately processed. This feature makes this camera type highly attractive for a digital photogrammetric measure- 
ment system which has to digitize concept models as fast as possible. The combination of such a photogrammetric 
measurement system with a CAD system appears as a turnkey system which makes CAD-based car design more 
flexible and significantly reduces production times. 
This paper reports on the new digital photogrammtric measurement system MATCH-I which is designed for the auto- 
matic digitization of industrial surfaces, and makes use of digital imagery. Especially the combination with a still video 
camera leads to a powerful measurement system, which can be efficiently integrated into a CAD system. The first part 
deals with the basic concepts and strategies and describes two fundamental measurement modes. The second part 
reports on the calibration of a still video camera KODAK DCS 200 based on a new bundle adjustment program pack- 
age. At the end, practical accuracy results and time statistics are given, captured from the measurement of a 20 % 
concept model and an interior clay model. 
2. BASIC CONCEPTS OF THE PHOTOGRAMMETRIC MEASUREMENT SYSTEM 
The strategic aims for the photogrammetric measurement system have to be as far-reaching as possible. They are 
addressed in simple terms as follows. The input for the system are digital images with natural or artificial texture and 
control points. 
(1): high absolute accuracy: € 0.1 - 2 mm 
(2): speed: 100 points/sec, full scale model in 3 days 
(3): dense surface digitization dependent on curvature 
(4): multiple photo matching strategy 
(5): robustness 
(6): system autonomy 
(7): self diagnosis 
(8): calibration procedure for non-metric cameras 
(9): characterline detection 
(10): integration into CAD system 
As already mentioned, the most important system features for a successful integration into a workflow environment 
are high accuracy (1) and speed (2), respectively. The system has to cover a wide accuracy range, since there are 
still users who require + 0.1 mm absolute accuracy, or even better. The system speed in terms of measurement rate 
should be in the order of 100 points per second. As far as car industry is concerned, it is desirable to measure a full 
scale model in 3 days including all process steps like object preparation, photo capture, digitization, point measure- 
ment, bundle adjustment and automatic surface reconstruction. 
A dense surface description (3), with curvature dependent point density seems important, since some users want to 
use such dense point clouds to mill the digitized physical model. Furthermore, a dense point cloud could also be the 
basis to analyse characteristic lines either in the measurement system in on-line mode, or by using visualization tools 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop “From Pixels to Sequences”, Zurich, March 22-24 1995