4-3-4 
Table 2 Results of cameras calibration 
Interior 
orientation 
Left Camera 
Right camera 
Value 
Standard deviation 
Value 
Standard deviation 
m x 
0.008169411 
5.9824897e-006 
0.0081828493 
6.7596162e-006 
m v 
0.0082745984 
6.0449712e-006 
0.0082854156 
6.8312629e-006 
353.97322 
4.0598375 
400.6065 
3.9840409 
y P 
240.72816 
3.2573169 
268.64533 
3.3235939 
a 
2.7212889e-005 
4.9883761e-005 
-2.9216815e-005 
4.7816647e-005 
K, 
0.00011596756 
0.00010720426 
-4.713952e-005 
0.00013371454 
K 2 
-2.0095956e-005 
1.7121584e-005 
6.543749 le-007 
2.2867503e-005 
K 3 
1.0714098e-006 
8.4285048e-007 
2.190873e-007 
1.209744e-006 
p, 
0.00011598028 
1.9949292e-005 
-0.00013673192 
1.9799675e-005 
P 2 
-3.8990547e-005 
1.732784 le-005 
-6.4237254e-005 
1.7044718e-005 
4 SYSTEM RELATIVE ORIENTATION 
The theoretical analysis and the results of relative orientation 
for images obtained during first technological mission have 
shown that the accuracy of simple relative orientation 
procedure (eliminating vertical parallaxes for corresponding 
points) results in unreliable estimation of relative orientation 
parameters. 
To obtain accurate relative orientation the following approach 
was applied. Firstly, to use spatial outdoor test scene for 
relative orientation; secondly, to apply special coded targets for 
precise subpixel targets location in the image. 
For estimation of relative orientation parameters the coplanarity 
condition was used: 
F = xx'(c{b{ - b{'c{) + x'f{t>2c{'- bi'c'i) + xz'(b{ci'- b^c{) + 
_ + Wi) + */'(%$' - - 
- x'zbfcj - zz'b^c'i - zf = 0 
b'l = - sin aB , b"l = - sin a'B cos co'B , 
b'2 = - cos aB , b"2 = cos a'B cos co'B , 
b'3 = 0, b"3 = - sin co'B , 
c'l= cos aB sin kB c'2= sin aB sin kB , 
c"l = cos a'B sin k'B - sin a'B sin co'B cos k'B , 
c"2 = sin a'B sin k'B + cos a'B sin co'B cos k'B , 
c'2= cos kB , c"3 = cos co'B cos k'B , 
a B , k b and a' B , co' B , k' b - elements of relative orientation, 
refering to the left and right images, correspondingly, x, z, f and 
x', z' /' - coordinates of points and focal distance of left and 
right images. To determine the elements of relative orientation 
the model of least squares method was used. 
To prepare outdoor spatial test scene 22 contrast coded targets 
were located at different distances from cameras in range of 10- 
25m. Targets positioning provides approximately equal density 
of targets in the images. 
The stereo images of the test scene in original software 
environment is shown in the Figure 4. 
where 
Figure 5. Stereo images of outdoor test scene for relative orientation using coded targets