4-3-5 
The developed coded targets [Knyaz, 1998] provide: 
5 SYSTEM EXTERIOR ORIENTATION 
• invariance to rotation and scaling 
• robust recognition of a target number with correction of 
possible mistakes 
• exact measurement of a target center 
• target detection in the image with non-uniform contrast 
• short processing time 
• compact size 
Elliptic approximation is used for high accuracy subpixel 
center of target measurement. The result of targets recognition 
is also shown in the Figure 4. 
To estimate relative orientation parameters bundle adjustment 
was performed using brief information of targets spatial 
location and estimating relative orientation parameters along 
with spatial coordinates of targets. It provides reliable and 
precise relative orientation approved during first mobile lab 
mission. 
Thus results of relative orientation obtained manually by 
operator from stero images of road for registration shown in 
■ Figure 6 gives 62.1 m estimation for distance between left 
j camera and fundament of bridge. The photogrammetric 
I measurement based on relative orientation, obtained by 
described above special procedure shows the value 84.8 m 
which is in good agreement with independent measurement - 
85 m. 
Performing relative orientation provides calculating 3D 
coordinates of the surveyed objects in the basis coordinate 
system related to left camera. But for 3D reconstruction based 
on the sequence of images external system of coordinate 
connected with road is needed. Choosing a set of coordinate 
systems for image sequence allows to define reference network 
and to determine common coordinate system. 
While mobile photogrammetric system is not equipped by 
exterior positioning system the exterior orientation is 
performed basing on road marks extraction. The possibility to 
choose reference points for orientation only in small local 
region of image causes low precision of estimation. To enhance 
accuracy of exterior orientation vanishing point (point of 
intersection of two lines in perspective image) is used. 
For the exterior orientation four corresponding points on the 
road marks are selected in the images, two (1,2) being on the 
left road mark and two (3,4) being on the right one. 
Coordinates of these points are used for calculating the 
coordinates of vanishing point (5) in the left and in the right 
images and then twelve orientation parameters are found by 
least-square estimation. 
The origin of the exterior coordinate system is placed to left 
marking line nearest point 1. The line 1- 2 gives an Y-axis of 
the exterior coordinate system. The point 3 gives the direction 
of the X-axis. Figure.5 demonstrates the interest points 
selection and vanishing points calculation. 
Figure 5. Road marks based exterior orientation using vanishing point 
The coordinates of interest points are calculated from relative For the test stereo pair shown in the Figure 5. the following 
orientatation and corrected by matching procedure. exterior orientation parameters were obtained (Table 3.):