The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
1075
P2
-0.005
0.0647
-0.0006
-0.0034
-0.0034
Y
0
0
0
0.0046
0
Table.l the coefficient of the camera model and the calibration results
experiment 1
experiment 2
experiment 3
experiment 4
experiment 5
image:
image:
image:
image:
3D calibration
fig-1, fig-2
fig-1, fig-2, fig.3
fig.3, fig.4
fig.3, fig.4, fig.5
result
fx
4093.1739
4092.8528
4093.4573
4091.1457
4092.9533
fy
4093.1739
4092.8528
4093. 4573
4091.1457
4092.9533
Uo
1613.6447
1613.5213
1599.6447
1593.4342
1612.9712
Vo
1203.3241
1209.7321
1201.3241
1200.8646
1210.3524
ki
-0.1196
-0.0232
-0.1196
-0.1260
-0.0231
k 2
0.0233
0.0183
-0.1196
0.0901
0.0045
k 3
0
0
0
0
0
Pi
-0.0011
-0.0009
-0.0011
-0.0009
0.0005
P2
-0.0012
-0.0001
-0.0012
-0.0011
0.0003
Y
0
0
0
0
0
Table.2 Photography model and the calibration results
The shape of the control points effect the results actually.
Because the centre of the shape is not the circle centre any more.
The photography model influent the results, too. In table 2, we
use the images in different photography ways.
When the model doesn’t cover the whole image, the result is
bad. When the model image is not in the centre of the image
(fig.5), the result is wrong!
At last, when the size of the object model is similar to the size
of the measurement object, the measurement precision is the
highest. You can see, fixed focal length, camera within a certain
distance of imaging is clear, because the cameras have a certain
depth of field.
Fig. 9 the modelling object
4. CONCLUSION
By experiments, we analyse the influence factors in the
calibration: the camera and lens, the shape of control points,
photography model and the scale between calibration object and
measurement. Appropriate model can get the right result
quickly, and the position of the model in the image effects the
calibration result directly. The control points should cover the
whole image evenly. The cross control point is better than circle
points. The more similar the size of object model and the
measurement object, the higher precision the calibration has.
Actually, each of them plays an important part in calibration.
We can get higher precision result based on considering those
influence factors. And the experiments in the end show that the
work is worthy. It means that we should take those factors into
account.
REFERENCES
Janne heikkila. Geometric Camera Calibration Using Circular
Control Points. IEEE Transactions on pattern analysis and
machine intelligence. Vol.22. No. 10. October 2000.
Triggs B. 1998. Auto-calibration from planar scenes. [A], In:
Proceedings of 5th European Conference on Computer Vision
[ C ], Freiburg Germany: 89-105.
Tsai RY. 1987, A versatile camera calibration technique for
high-accuracy 3-D machine metrology using off-the-shelf TV
cameras and lens. IEEE Journal on Robotics and Automation,
RA-3(4): 323-344. [J]
Yongjun Zhang, 2002. Camera Calibration Using 2D DLT and
Bundle Adjustment with Planar Scenes, Geomatics and
Information Science of Wuhan University, Vol.27 No.6.
Zhengyou Zhang, A Flexible New Technique for Camera
calibration. December 2, 1998.