Boersma, Saskia M. 
Furthermore, no reference information is required during image acquisition. No reference frame or scale bar is required 
when the wound images are made. The cameras are mounted, with a fixed position, in a triangular. frame. The 
calibration process determines the position of the cameras in the frame. The software for the DSM generation uses this 
information as exterior orientation data. 
The system should be as flexible as possible for the use in a hospital environment. This implies that it should be as light 
and small as possible. Taking into account the requirements described above, we designed the following vision system. 
1.2 Final design 
The three progressive scan video cameras are mounted on a triangular frame with a light source and texture proj ection 
in the centre (Figure 2). The distance between the cameras is fixed to 15 cm. The base-to-height ratio of 1:2 implies that 33 
the distance from the cameras to the wounds has to be approximately 30 cm. The cameras are mounted in a slightly 
convergent way. The pa 
NE determ: 
STE Side-view of camera : errors ( 
automa 
in Z. A 
object : 
  
  
Side-view 
of frame 
  
  
  
Back-view 
of frame 
  
   
  
  
  
  
  
  
  
  
  
  
  
I 
HAITI 
Camera 2 Camera 3 
  
  
  
   
Figure 2: Camera configuration in frame Figure 3: Picture of system configuration 
(illumination through the hole in the center) 
  
3 CALIBRATION 
  
  
  
  
  
  
  
  
  
3.1 Purpose of calibration We no 
provid 
The cameras we used are non-metric and not designed for high accuracy purposes. This makes the calibration of the or 
system decisive for the accuracy of the DSM. The calibration consists of two different parts: the calibration of the 
interior orientation including lensdistortions of the individual cameras and the calibration of the frame in which the Table 
exterior orientation elements of the three cameras are determined. Similai 
A complete camera calibration includes radiometric aspects (e.g. the compensation of different sensitivity of CCD Gne T 
sensor cells) as well as geometric aspects. No radiometric camera calibration is performed. The geometric camera 
isis (d one ; : ; about t 
calibration involves the determination of the focal length, the location of the principal point and the parameters of the System 
lens distortion. The software package Bingo (Kruck, 1998) is used. This package enables us to determine calibration y 
parameters for more than one camera simultaneously. 
3.2 Calibration set-up and measurement Cam 
Foca 
We calibrated the system using a testfield of 30 x 40 cm with 63 targets in a regular grid with a mutual distance of 9 cm. Princ 
Six targets were placed 8 cm above the plane to establish a 3D configuration of the targets. The circular targets are Princ 
black on a white background with a diameter of 0.5 cm (Figure 4). A total of 8 image triplets (figure 5) were made, 4 | Para: 
convergent and 4 normal case scenes according to Wester-Ebbinghaus (1983). This results in a total of 24 images. Parai 
Every target was imaged in average in 18 images. We applied template matching for subpixel image measurement. The Para: 
co-ordinates of the image points were measured with template matching with an average precision of 3 x 10 mm, Parar 
corresponding to an precision of 5 % of a pixel. Parar 
  
  
  
86 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000.