The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
396
represents the airborne LiDAR data. Terrestrial data covers
very narrow areas compared to the airborne data, but the point
density is much higher. Even though small areas are selected
from the terrestrial data for the process, the process can take
some time, and one should give attention to the lack of a
computer memory for computation. Both data were captured in
the same area, but the overlap area of the two systems is not
large.
Figure 11. Terrestrial laser scanning data (a) is used as
reference data to calibrate airborne laser scanning data (b)
AX AY AZ Aco A $ Ak
(m) (m) (m) (deg) (deg) (deg)
Given
0.210 0.190 -0.003
0.168
-0.035
0.000
parameters
Normfml : 0.159
N/A
ICP
-0.163 -0.080 -0.087
0.068
-0.052
0.118
Norm[m] : 0.119
0 : 0.113
ICPatch
-0.032 -0.015 0.026
0.009
-0.012
0.018
Norm[m] : 0.139
G : 0.125
Planar
0.108 -0.495 -0.003
0.174
0.040
-0.032
Patches
Norm[m] : 0.129
0 : 0.120
Table 3. The LiDAR system calibration test using real data (6
bore-sighting parameters)
Table 3 shows the adjusted bore-sighting parameters from the
proposed methods, respectively. To evaluate the adjusted
parameters, average normal distances between the adjusted
surface and reference surface are calculated. After the object
surfaces are re-constructed using the new system parameters, a
surface matching procedure was then carried out to find
corresponding points between both data; ICPatch was used for
this purpose, in this case. Consequently, normal distances
between matched points and triangular patches are calculated.
The real LiDAR data was captured along the rail-road areas in
eastern Canada, and these areas are quite rural. Hence, it was
hard to extract the planar patches, especially from man-made
objects such as buildings. Since the distribution and
configuration of control patches are important in terms of
possible correlations between calibration parameters, one
should give attention to the extraction of control data. From the
control feature selection point of view, the other two
approaches, using ICP and ICPatch, appear easier and more
effective. These two methods, however, are very sensitive to the
initial approximations and random error size. Even though
pseudo-conjugate points from the ICP procedure and triangular
patches from TIN are easier approaches in terms of establishing
corresponding points for the rural areas, like this test area, the
method using segmented planar patches can have reliable
solutions and is not sensitive to the ill conditioned data like
high random errors, if planar patches can satisfy the required
condition; configuration and distribution. For these reasons,
large and abundant planar patches are relatively better than the
closest points and the closest TIN element; which are not very
sensitive to random errors and initial approximations.
Furthermore artificial control planar targets and well-known
man-made objects can be considered as the ideal control data
for the system calibration.
Figure 12. (a) the TIN represents the reference data and points
denote target data in 2D display, (b) in 3D display, reference
data (terrestrial LiDAR) is mainly appeared along the vertical
wall, while the points of the target data (airborne LiDAR) are
very dense on the ground.
4. CONCLUSIONS
In this paper, the author introduces the airborne LiDAR system
calibration procedure using the terrestrial LiDAR data which is
capture in the same area. Because terrestrial LiDAR systems
usually have shorter ranges and much higher point density,
those object surface data works well for the airborne LiDAR
system as reference data. Three approaches are used for
extracting conjugate features; pseudo-conjugate points by ICP,
conjugate poin'ts/triangles by ICPatch, and conjugate planar
patches by plane segmentation. And the real data test shows
that existing bore-sighting parameters are improved after
calibrating system using LiDAR raw measurement, which is
confirmed by calculating the normal distances between
reference surfaces and adjusted surfaces. For increasing the
robustness and reliability of the LiDAR system calibration,
strong surface match procedure should be also considered in the
future.
REFERENCES
Baltsavias, E., 1999. Airborne laser scanning: existing systems
and firms and other resources, ISPRS Journal of
Photogrammetry and Remote Sensing, 54 (2-3): 164-198.
Bretar F., M. Pierrot-Deseilligny, and M. Roux, 2004. Solving
the Strip Adjustment Problem of 3D Airborne Lidar Data.,
Proceedings of the IEEE IGARSS’04, 20-24 September,
Anchorage, Alaska.