1.1 Hardware System 
The dot projection process excels in areas of dynamic 
measurement the system described in this paper was designed 
for real-time measurement. The basic criteria of the system was 
to measure and model both stationary (static) structures as well 
as nonstationary (dynamic or deploying) structures at a frame 
rate of at least 900 frames per minute for a continuous time 
period of at least 15 minutes. The 3d object point coordinates 
are obtained by two synchronized instrumentation grade digital 
cameras each interfaced to a high-speed digital frame grab 
board for bandwidth. The system is designed to allow for multi- 
processor, multi disk system that will support multiple camera 
units, that is interfaced to a custom projection system. The 
projectors are configured to operate in different environments. 
The projector used in this study can operate in a harsh 
environment, primarily for this case to work inside a vacuum 
  
Figure 4 Projector 
The projector can be configured with several dot pattern sizes, 
the least will display up to 4000 dots. The system can be 
configured in several ways real time of buffered frame store 
then read. The buffer frame/disk store method was configured 
for this system due to the 15 frame per second data acquisition 
from the two cameras. The points on the left and right image 
are automatically windowed and read. The software utilizes a 
multi target extraction tool for measurement. Prior to 
automated measurement the registration of the two cameras 
must be made to determine camera position and orientation. 
One method is to use a target template this provides for 
automated calibration and camera registration with the target 
extraction tool. 
° ° 
° e 
e $ 
. ° ° . 
  
Figure 5 Calibration Registration template 
1.2 Automated Measurement 
The stereo matching principal used in ShapeCapture and 
ShapeMonitor software is based on epipolar geometry as 
shown in Figure 6. 
The object point P, the two projection centers of the two 
images, O1 and O2, form a plane (the epipolar plane) that 
intersects the two image planes. As a result, two epipolar lines 
are created. The matched points fall on these lines. Once the 
camera positions and orientations at the two image locations 
are known, the equations of the epipolar lines become known. 
We will now start by taking the coordinate pl in the left image 
and look for point p2 on the epipolar line in the right image. 
Since there could be other points falling on the same line, we 
need additional constraint to make sure we select the 
appropriate point. 
N 
e 
b 
YT zi 
  
  
  
Ÿ 
Figure 6 Epipolar Model 
Once the calibration and registration is set the digital image 
frames from the two cameras are dumped into the memory 
buffer, id coded and saved on disk. The program automatically 
opens the coded files, the thousand of targets extracted, stereo 
matched, and coordinates computed and the file closed in 
approximately 2 seconds then the next set is opened. The 
program data acquisition module can be configured for the 
overall number of frames and run unattended. 
The system was original designed for measurement of thin skin 
membranes for deployment inside a vacuum chamber but it 
works well on other surfaces including static subjects. 
2.0 THIN MEMBRANE SURFACES 
Measurement of thin membrane surfaces presents problems 
using previous photogrammetric techniques such as retro 
reflective targeting. The surface material can sometimes be so 
thin that touching the surface can damage the material 
especially in placement of targets. Also if the object is 
compacted and is being deployed (inflated) over a period of 
time it would be impossible to monitor the targets during 
deployment. Figure 7 illustrates a thin membrane where 
wrinkle characteristics are being monitored. Figure 8 displays 
the modeling of the skin surfaces for deviations and wrinkles. 
-80— 
  
In ad 
conv 
data : 
probl 
occui 
to wc 
dispe 
Figur 
thin ı 
mour