The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
LiDAR system calibration is done through several procedures.
In the first step, individual sensors are calibrated in the
laboratory. After that, mounting parameters are determined
after installing sensors on a platform. In the last step, in-situ
calibrations are done before and after surveying missions
(Schenk 2001).
Platform
Figure 1. Coordinates and parameters involved in a LiDAR
acquisition system
X
— Xq + R(x)j>icT >+ R(o<t>K R&a>Aj>AicR/3
0
0
-P
(1)
1,2 Previous Researches
In the past few years, several research methods for LiDAR
calibration have been developed. For example, Schenk (2001)
investigated the sources of systematic errors that can occur in a
LiDAR system. A calibration procedure was then proposed
using such an analysis. This work revealed that some of the
calibration parameters cannot be easily estimated due to their
strong correlation. The calibration methodology developed by
Morin (2002) uses the LiDAR equation to solve for the bore
sighting misalignment angles and the scanner angle correction.
These parameters are either estimated using ground control
points or by observing discrepancies between tie points in
overlapping strips. However, the identification of distinct
control and tie points in LiDAR data is a difficult task due to
the irregular nature of the collected point cloud. To alleviate
this difficulty, Skaloud and Lichti (2006) presented a
calibration technique using tie planar patches in overlapping
strips. The underlying assumption of this procedure is that
systematic errors in the LiDAR system will lead to non
coplanarity of conjugate planar patches as well as bending
effects in these patches. The calibration process uses the
LiDAR equation to simultaneously solve for the plane
parameters as well as the bore-sighting misalignment angles.
However, this approach requires having large planar patches,
which might not always be available. In addition, systematic
biases, which would not affect the coplanarity of conjugate
planar patches, could still remain. The approaches taken by
LiDAR surveying companies were more closely applicable in
practice. For example, to calibrate the LiDAR system, Hanjin
(2006) devised a calibration field which is composed of well-
known surfaces. Using the calibration site, discrepancies
between the LiDAR point cloud and the reference surface are
observed and used to determine the system parameters such as
the bore-sighting roll and pitch angles and scale parameters.
The drawbacks of this approach are that the method involves
manual and empirical procedures, and some parameters are not
considered in the calibration procedure.
This paper presents a new methodology for simultaneous
estimation of the LiDAR bore-sighting parameters using control
features that are automatically extracted from a reference
control surface. In this approach, the reference control surface
is derived from a terrestrial LiDAR system. The shorter ranges
and the high point density associated with terrestrial LiDAR
systems would ensure the generation of a reference surface,
which is accurate enough for reliable estimation of the
calibration parameters associated with airborne LiDAR systems.
2. PROPOSED METHODS
2.1 Point Primitives and ICP method
The main issue considered in the proposed calibration
procedure is identifying conjugate features in the airborne and
terrestrial LiDAR systems. As mentioned above, due to the
irregular nature of the generated point cloud from a LiDAR
system, it is usually believed that there is no point-to-point
correspondence between the derived points from the airborne
and terrestrial systems. However, one might argue that point-to-
point correspondences for a fraction of the terrestrial and
airborne datasets can be assumed; considering the higher point
density associated with terrestrial systems compared with that
for airborne systems and the noise level in both datasets.
Therefore, this paper introduces a point-based calibration
procedure using pseudo-conjugate points in the terrestrial and
airborne datasets, and Equation 2 shows the target function
based on the point primitives for the calibration. In this
Equation, the superscription T denotes the target data (airborne
LiDAR data), while R denotes the reference data (Point cloud
generated by terrestrial LiDAR system). While only point cloud
coordinates are utilized from the reference data, LiDAR system
raw measurements should be available for the target system. As
shown in
Figure 2, two points are used from both datasets, and
points in reference data*
points in calibration target data*
Figure 2. The closest points selected by the iterative closest
point procedure
those points are selected by the iterative closest point (ICP)
procedure (Zhang, 1994), which is one of the common surface
matching method. After that, in the calibration procedure, two