4-3-3 
3 CCD CAMERAS CALIBRATION 
To obtain image scale suitable for accurate 3D scene 
reconstruction CCD cameras is equipped with 25 mm focal 
length lens. Their interior orientation parameters were 
determined by calibration procedure in laboratory conditions. 
The plane wall with 25 contrast targets was used as testfield 
(Figure 3.), the dimensions of the testfield being about 1.5m x 
1.2m. It provides approximately equal targets density in 
captured images at infinity adjustment.. 
S un der vibration 
e prove d stable I 
' r loads occurring I 
The contrast targets were used as reference points for 
calibration. At this stage of system development there were no 
means to measure spatial coordinates of reference points with 
high precision. So they have to be estimated during the 
calibration. 
Calibration stages 
Initial values definition. The focal length / was taken as 25 
mm. The principal points coordinates were taken as it lays in 
the center of the CCD frame. The CCD image scales in X and 
Y directions respectively were taken as m x =m y =0.008 mm/pix. 
Image coordinates determination for reference points. For 
each captured image coordinates of reference points was 
estimated in semi-automatic mode. For subpixel accuracy 
weight operator was applied. 
Images exterior orientation. For each image exterior 
orientation parameters in coordinate system of the testfield 
were determined, object reference points coordinates and 
internal orientation parameters being frozen. 
lodes: 
rames in the true 
sequences for 1 
1 P* x ds images at 
innat with YUV 
the program is 
; v ideo" (overlay) 
°n the monitors. 
a single frame or 
is brought back 
Reference distances measurements. While testfield reference 
coordinate were known with low accuracy the reference 
distances between some of the reference points were measured 
with higher precision for better system determinability. The 
distances used for calibration are presented in Table 1: 
Table 1. Reference distances used for calibration 
Reference point # 
Distance, mm 
1-25 
1561.3 
5-21 
1560.5 
5-25 
998.5 
1-21 
998.5 
1-5 
1200 
21-25 
1200 
Figure 3. Testfield for the calibration procedure 
For calibration procedure 24 images of testfield were captured 
(12 images for each camera). The average distance between 
camera and testfield was approximately 7 m. The angle position 
of cameras with respect to the testfield was: 
a«0,±45°, (o*0,±45°, k*0, 180° 
Targets image coordinates were determined with subpixel 
accuracy with weight operator. The additional parameters 
describing real physical system in collinearity conditions were 
taken in form: 
Ax = Ax p + ya + xr 2 K x + xr A K 2 + xr 6 K 2 + (r 2 + 2x 2 )P X + 2 xyP 2 
Ay = Ay p +xa + yr 2 K x + yr*K 2 + yr 6 K 2 + 2xyP x + (r 2 + 2 y 2 )P 2 
x = m x (x-x p );y = -m y ( y - y p );r = yjx 2 + y 2 
The parameters of interior orientation: principal point Xp, yp, 
scales in x and y directions m x , m y , and affinity factor a, the 
radial symmetric K,,K 2 ,K 3 distortion and decentering P/,P 2 
distortion are to be estimated. 
Design of estimated parameters vector. The following 
components were included in estimated vector: 
• exterior orientation parameters of all the images 
• interior orientation parameters of both cameras 
• coordinates X, Y of all the object reference points 
excluding the central point (#13) and X coordinate of the 
point #5, which determine the object space coordinate 
system. Z coordinates of all points was taken as 0. 
Estimation of unknown parameters. This stage of calibration 
was performed by the original bundle adjustment software 
developed for Windows 95. 
Results of estimation interior orientation parameters are shown 
in the Table 2. 
The accuracy of estimated unknowns was: 
• Standard deviation of spatial reference points coordinates 
a xs = 0.081 mm 
• Standard deviation of angle exterior orientation parameters 
«W* 0.065° 
• Standard deviation of residuals of collinearity conditions 
CTio= 0.260 mm