racy 
me- 
itivi- 
out- 
T of 
ann, 
uali- 
irio- 
hree 
mul- 
levi- 
rved 
Fthe 
ngly 
'CD 
Orts 
CD 
itial 
era. 
with 
ield 
ure- 
car- 
lgo- 
por- 
oto- 
IS is 
the 
0 in 
2. SELFCALIBRATION OF THE PROGRES 3000 
2.1 The ProgRes 3000 camera 
  
The ProgRes 3000 camera /Lenz 1989/ is a digital CCD 
camera with a sensor chip of 512 * 580 pixels at a pixel 
size of 17 um * 11 um. High resolution images are 
generated from a number of successively taken images, 
so called partial images. For each partial image the CCD 
sensor is moved a fraction of the distance between two 
pixel centres using a piezo-controlled aperture displace- 
ment (PAD). The amount of this displacement, and thus 
the image resolution is programmable and can be per- 
formed with a root mean square error of about 0.2 m 
/Lenz, Lenz 1990/. This microscanning procedure re- 
sults in a resolution of up to 3072 * 2320 pixels and a pixel 
size of down to 2.83 um * 2.75 um. Additionally, colour 
imagery can be captured. At maximum resolution the 
colour image acquisition can be performed in a few 
seconds. It has been shown that digital images of static 
objects can be acquired with the same image quality as 
hardcopy diapositives. 
2.2 The mathematical model 
  
Selfcalibration is performed based on the well known 
collinearity equations. Corrections for the improvement 
of the coordinates of the principal point and the prin- 
cipal distance are introduced. Additional correction 
terms dx and dy for the measured image coordinates x 
and y respectively are formulated to model lens distor- 
tion. Higher order polynomials can be used to correct 
for other systematic image deformations from the cen- 
tral projection. 
If parameters for radial and tangential lens distortion are : 
to be determined, dx and dy can be expressed as: 
dx 2 Ai() - r)x * As(r! - r$)x 
+ B1(y° + 3x") + 2B2xy 
dy=A (7 —R5)y+A2(r" —rÉ)y 
+2B,xy+B2(xX"+3y) 
with: 
Xy image coordinates 
4Ai,42,r, parameters for radial lens distortion 
Bı,B2 parameters for tangential lens distor- 
tion 
P-(x-x)-*(y-»y 
Xo , yo coordinates of principal point 
From the observations (image coordinates of a number 
of tie points) the estimation of the unknown parameters 
(corrections for the coordinates of the principal point 
and the principal distance, A1 , A2 , B1 , B , the elements 
of exterior orientation and the object coordinates of the 
tie points) is performed according to the least squares 
principle. Control information can be introduced in ad- 
dition to the image coordinates in order to define the 
elements of the datum (three translations, three rota- 
tions and scale). 
2.3 Image acquisition 
For the geometric calibration of the ProgRes 3000 a 
relatively small three dimensional testfield with 29 
equally spaced points was used. The size of the testfield 
is approximately 0.60 m * 0.60 m, the maximum height 
difference is 0.23 m. The signals are black circles on a 
white background and have a diameter of 3 mm each. 
The coordinates of the 29 points were determined 
photogrammetrically using a Wild P31 metric camera 
and a Rollei Réseau Scanner. The resulting standard 
deviations of the control point coordinates are about 
12 um in X, Y, and 25 um in Z. 
Eight images of the maximum resolution of 3072 * 2320 
pixels were taken with the ProgRes 3000 equipped with 
a standard 16 mm lens at a distance of about 1.6 m, 
yielding an image scale of about 1:100. Thus, one signal 
covered about 10*10 pixels. Four of the eight images 
were. oblique views taken from an angle of about 50 
  
Figure 1: Geometric configuration of image acquisition