The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
204
overlapping LiDAR strips and control surfaces. Due to the
irregular nature of the LiDAR footprints, the identification of
distinct points (e.g., building comers) is quite difficult and not
reliable. More suitable primitives have been suggested by Maas
(2000), where the correspondence is established between
discrete points in one LiDAR strip and TIN patches in the other
one. The correspondences are derived through a least-squares
matching procedure where normal distances between conjugate
point-patch pairs are minimized. The drawback of this work is
that simple shifts were used as the transformation function
relating conjugate point-patch pairs. The validity of such a
model was not completely justified. Moreover, the estimated
shifts were not used to derive an indication of the point cloud
quality. Bretar et al., (2004) proposed an alternative
methodology for improving the quality of LiDAR data using
derived surfaces from photogrammetric procedures. The main
disadvantage, which limits the practicality of this methodology,
is relying on having aerial, imagery over the same area. In
addition, the proposed approach uses an affine transformation to
relate LiDAR and photogrammetric surfaces without sufficient
justification. In Pfeifer et al. (2005) and Vosselman (2004)
other methods were developed for detecting discrepancies
between overlapping strips. Detected discrepancies were used
for strip adjustment procedures rather than system and data
evaluation.
The objective of this paper is to propose a cost-effective and
meaningful quality control (QC) method, which is based on
analyzing the compatibility of LiDAR data in overlapping strips.
More explicitly, the objective of the presented research is to
develop a tool for detecting the presence of systematic biases as
well as inspecting the noise level in the point cloud with a
satisfactory level of automation (i.e., requiring minimum user
interaction). The paper will start with a brief discussion of the
LiDAR mathematical model relating the system measurements
to the ground coordinates of the point cloud. Then, an analysis
of possible systematic and random errors and their impact on
the resulting surface will be outlined. Following the discussion
of the error sources and their impact on the accuracy of the
LiDAR footprints, a QC tool will be proposed. The paper will
conclude by experimental results from a real dataset involving
overlapping strips from operational LiDAR system. The results
have proven the feasibility of the introduced methodology to
evaluate the quality of the LiDAR data. More specifically, the
proposed measure detected the presence of systematic biases
and the data noise level. Future research will focus on relating
the detected discrepancies between overlapping strips to the
system biases.
2. LIDAR MATHEMATICAL MODEL
The coordinates of the LiDAR footprints are the result of
combining the derived measurements from each of its system
components, as well as the bore-sighting parameters relating
such components. The relationship between the system
measurements and parameters is embodied in the LiDAR
equation (Vaughn et al., 1996; Schenk, 2001; El-Sheimy et al.,
2005), Equation 1. As it can be seen in Figure 1, the position of
the laser footprint, X G , can be derived through the summation
of three vectors (X o , P G and /5 ) after applying the appropriate
rotations: R . . „ R. . , . and R In this equation,
yaw, pitch, roll ’ Aco,A</>,Ak ct,p ^ 9
X o is the vector between the origins of the ground and IMU
coordinate systems, P is the offset between the laser unit and
IMU coordinate systems (bore-sighting offset), and p is the
laser range vector whose magnitude is equivalent to the distance
from the laser firing point to its footprint. The term
Ryan pitch roil stan ds for the rotation matrix relating the ground
and IMU coordinate systems, R A(o A(A Av represents the rotation
matrix relating the IMU and laser unit coordinate systems
(angular bore-sighting), and R a p refers to the rotation matrix
relating the laser unit and laser beam coordinate systems with
a and fd being the mirror scan angles. For a linear scanner,
which is the focus of this paper, the mirror is rotated in one
direction only (i.e., a is equal to zero). The involved
quantities in the LiDAR equation are all measured during the
acquisition process except for the bore-sighting angular and
offset parameters (mounting parameters), which are usually
determined through a calibration procedure.
3. LIDAR ERROR BUDGET
The quality of the derived point cloud from a LiDAR system
depends on the random and systematic errors in the system
measurements and parameters. A detailed description of LiDAR
random and systematic errors can be found in Huising and
Pereira (1998), Baltsavias (1999), and Schenk (2001). The
magnitude of the random errors depends on the accuracy of the
system’s measurements, which include position and orientation
measurements from the GPS/IMU unit, mirror angles, and
ranges. Systematic errors, on the other hand, are mainly caused
by biases in the bore-sighting parameters relating the system
components as well as biases in the system measurements (e.g.,
ranges and mirror angles). In the following sub-sections, the
impact of random and systematic errors in the system
measurements and parameters on the reconstructed object space
will be analyzed.
3.1 Random Errors
The purpose of studying the impact of random errors is to
provide sufficient understanding of the nature of the noise in the
derived point cloud as well as the achievable accuracy from a
given flight and system configuration. In this work, the effect of
random errors in the system measurements is analyzed through
a simulation. The simulation process starts from a given surface
and trajectory, which are then used to derive the system
measurements (ranges, mirror angles, position and orientation
information for each pulse). Then, noise is added to the system
measurements, which are later used to reconstruct the surface
through the LiDAR equation. The differences between the
noise-contaminated and true coordinates of footprints are used
to represent the impact of a given noise in the system
measurements. The following list summarizes the effect of
noise in the system measurements.
• Position noise will lead to similar noise in the derived
point cloud. Moreover, the effect is independent of the
system flying height and scan angle.
• Orientation noise (attitude or mirror angles) will affect the
horizontal coordinates more than the vertical coordinates.
In addition, the effect is dependent on the system flying
height and scan angle.