COS K' p; 
COS K'p; 
for the left 
ation for the 
stance of left 
] distance of 
the model of 
on procedure 
ever cameras 
ides a set of 
number and 
; taken from 
f the left and 
m. Figure 6 
ring relative 
ve orientation 
ets. 
  
  
    
RR 
orientation 
pratory vehicle 
nately 30 m in 
is taken. Then 
nd next some 
) while vehicle 
moving. After then reference point are recognised in all 
acquired images and their image coordinates are found with 
sub-pixel accuracy. Then bundle adjustment procedure is 
performed, estimated vector of unknowns including all 
reference points coordinates and relative orientation parameters 
in condition of given reference distances and co-planarity of 
reference points for each stereo pair. 
Table 7 represents given reference point distances for relative 
orientation and residual errors in reference distances after 
bundle adjustment. Two reference distances are used for 
orientation between points #6 — #9 and points #11 - #12. The 
rows 3, 4 of Table 7 correspond to second image acquisition car 
position and the rows 5, 6 of Table 7 correspond to the third 
image acquisition car position. 
  
  
  
  
  
  
  
  
  
  
  
  
# Reference Distance, m Residuals, 
point # m 
] 6-9 1.24 0.000 
2 11-12 1.6 0.001 
3 1006-1009 1.24 0.000 
4 1011-1012 1.6 -0.001 
5 2006-2009 1.24 0.001 
6 2011-2012 1.6 0.002 
  
Table 7. Reference distances used for relative orientation and 
residual errors in reference distances after bundle adjustment 
The results of relative orientation for concerned image set are 
given in Table 8. 
  
  
  
  
  
  
  
  
  
  
Relative Value, ° Residual errors, ° 
orientation 
parameter 
Op 1.603004 6.9165473e-005 
Kp 0.48137715 -3.5476828e-006 
Œ'B 1.5635437 6.8559054e-005 
Q'5 0.2256283 -3.3246774e-006 
K’ 0.03901096 -4.6470541e-008 
  
Table 8. Relative orientation parameter estimations 
The estimated relative orientation parameters are transmitted to 
the obstacle detection program and used for road 3D model 
reconstruction based on lane markings detection and for 
obstacle parameters determination. 
5. RESULTS OF SYSTEM APPLICATION 
The developed method for calibration and relative orientation of 
automobile based photogrammetric system was investigated in 
a tested area and in real high-way conditions. For the 
investigation in a tested area a white box of known sizes was 
used. The measurements of obstacle distance from the 
laboratory vehicle were made by obstacle detection system 
based on calibration results and by independent means (by tape- 
measure). The results of investigations show that proposed 
calibration technique provides distance estimation with 
accuracy of 0.1 m at obstacle distance of 20 m and about 1 m at 
obstacle distance of 100 m. 
  
Figure 9. A sample of obstacle parameters estimation 
A sample of obstacle detecting and obstacle parameters 
estimation is presented in Figure 9. The results of obstacle 
detection system working are shown in the image. The 
estimated distance is 82.7 m, obstacle width is 0.5 m, the 
distance from the obstacle to the left marking line is 0.9 m and 
the distance from the obstacle to the right marking line is 1.1 m. 
These results are in good agreement with data of independent 
measurements. 
6. CONCLUSIONS 
The proposed method for vehicle stereo system calibration 
provides high accuracy of 3D measurements even in bad stereo 
condition caused by restriction of automobile geometry. This 
fact was approved during development and testing of obstacle 
detection software. The results of system laboratory testing 
show that proposed calibration technique provides distance 
estimation with accuracy of 0.1 m at obstacle distance of 20 m 
and about 1 m at obstacle distance of 100 m. Achieved accuracy 
is quite enough for the tasks of obstacle detection and obstacle 
parameters estimation. 
The developed method has high degree of automation due to 
applying coded targets as reference points markers and because 
of simplicity of a plane test field used for calibration. The 
calibration procedure could be fully automated for industrial 
purposes by using rotating stages with 2 degrees of freedom. 
REFERENCES 
Knyaz V.A., Sibiryakov A. V.., 1998. The Development of 
New Coded Targets for Automated point Identification and 
Non-contact 3D Surface Measurements, International Archives 
of Photogrammetry and Remote Sensing, Vol. XXXII, part 5, 
Hakodate, Japan, 1998, pp. 80-85. 
Knyaz V.A., Zheltov S.Yu., Stepanyants D.G., 1999. Method 
for accurate camera orientation for automobile 
photogrammetric system. Proceedings of International 
Workshop on Mobile Mapping Technology, Bangkok, Thailand, 
April 21-23,1999, pp.4-3-1 — 4-3-6 
Stepanyants D.G., Knyaz V.A., 2000, PC-Based Digital Close- 
Range Photogrammetric System for Rapid 3D Data Input in 
Cad Systems, International Archives of Photogrammetry and 
Remote Sensing, Vol. XXXIIL, part B5/2, Amsterdam, The 
Netherlands, 2000, pp. 756-763 
—583—