Boresight Calibration of
Low Point Density Lidar Sensors
Sudhagar Nagarajan and Shahram Moafipoor
Abstract
Mobile Mapping is the technique of acquiring accurate
geospatial information of a scene using multiple sensors
mounted on a moving platform. At the core of these sys-
tems is the direct georeferencing techniques that tie together
multi-sensor data on-board. An important aspect of direct
georeferencing is to apply accurate boresight calibration of
individual sensors with respect to the platform body frame.
Conventional techniques use Ground Control Points (
GCP
)
for this calibration. Considering the challenges in identifying
GCPs
from low density lidar point cloud, this research pres-
ents a feature-based registration method that uses control
planes. The presented method is performed in a lab-facility
utilizing static data to determine the alignment between
platform body frame and lidar frame by minimizing the
volume formed between low point density lidar and control
planes. The paper discusses the mathematical models and
feasibility of the technique for use in mapping applications.
Introduction
Recent advances in lidar sensor technologies have made
low-cost and light-weight Mobile Mapping Systems (
MMS
)
available in the market. Examples of such
MMS
include man
portable and
sUAV
(small Unmanned Aerial Vehicle) based
lidar systems. These
MMS
are mainly used for asset inventory,
as-built survey, emergency response and landuse mapping
applications where expected accuracy is less as compared to
survey-grade
MMS
. A typical
MMS
consists of multiple posi-
tioning and mapping sensors. The positioning sensor include
Global Navigation Satellite System (
GNSS
) receiver, Inertial
Measurement Unit (
IMU
) and Distance Measurement Instru-
ment (
DMI
). Lidar scanner and camera are the most common
mapping sensors used in the
MMS
. The overall accuracy of the
MMS
system is mostly dominated by the quality of the
GPS/IMU
solution utilized in generating a trajectory of the mobile plat-
form. Thus, for the highest accuracy, the integrated
GPS/IMU
unit utilizes Real-Time Kinematic (
RTK
) or Post-Processing
Kinematic (
PPK
) differential corrections. After aligning and
time-synchronizing sensory data, the multi-sensor data can be
georeferenced. In a
MMS
, the positioning and mapping sensors
are mounted physically apart from each other on the mobile
platform. In order to perform multi-sensor data georeferenc-
ing, the alignment (rotation and translation) between the
sensors need to be determined. The process of determining
the spatial alignment of lidar sensor and other mapping sen-
sors with respect to
IMU
body frame is referred as boresight
calibration.
Background
Low-cost and portable
MMS
utilize sensors similar to mapping
grade systems except they are characterized by low resolution
and accuracy. While lower weight and costs are desirable,
the commensurate lower resolution and accuracy are not.
The lower resolution can typically be compensated by repeat
coverage. However, compensating lower accuracy is more
challenging. The major factors that must be taken in account
are stable boresight calibration parameters, quality of
IMU
and lidar,
GNSS
signal reception and ground control informa-
tion to model the systematic errors. The other limitation in
utilizing low-cost/weight lidar sensors is their limited vertical
field of view. Hence, the inclination of the lidar may need
to be changed depending on the object of interest. After each
change, the system needs recalibration so accurate boresight
parameters can be determined.
Traditional techniques determine boresight parameters in
static or in-flight mode using control points. However, iden-
tifying control points in a low-resolution lidar point cloud is
nearly an impossible task (Wehr and Lohr, 1999). Also, the
uncertainty in choosing control points can significantly affect
the accuracy of derived point cloud data (Jozkow
et al.
, 2016).
Commercial off-the-shelf
MMS
usually comes with known
boresight parameters, whereas, custom-built
MMS
requires a
boresight calibration procedure to be performed by the user.
Furthermore, considering the limitations on resolution and
viewing angle of the custom-built, low-cost
MMS
, the align-
ment of the sensors may need to be adjusted often which re-
quires frequent boresight calibration. The temporal instability
of lidar sensor parameters also warrants frequent calibration
of boresight parameters (Le-Scouarnec
et al
., 2013).
The geometry of direct georeferencing is illustrated in
Figure 1 where coordinate frames of each sensor in a
MMS
is
shown. Sensors typically record data in their local coordi-
nate system. Hence, it is necessary to transform the recorded
measurements to a global coordinate system to perform direct
georeferencing. For this reason, the data recorded by lidar
sensors need to be transformed from their local sensor frame
to a map frame. The coordinate system of the laser scanner
is denoted by axes X
L
, Y
L
, and Z
L
in the figure. All angle and
range measurements by the laser scanner are with respect to
this coordinate system that is often referred as the
SBF
(Sen-
sor’s Body Frame). Equation 1 transforms the measured fea-
ture coordinates from
SBF
to
IMU
body frame whose origin and
axes are represented by X
b
, Y
b
, and Z
b
in Figure 1. These
IMU
coordinates of objects in the scene are further transformed to
ground coordinate system in the a-frame (arbitrary), e.g.,
ECEF
(Earth Centered and Earth Fixed) or a local map frame.
P
a
=
P
b
a
+
R
n
a
R
n
b
(
R
L
b
p
k
+
Δ
T
L
b
)
(1)
Sudhagar Nagarajan is with the Florida Atlantic University,
Department of Civil, Environmental and Geomatics
Engineering, 777 Glades Road, Boca Raton, FL 33433
(
).
Shahram Moafipoor is with Geodetics, Inc., 2649 Ariane
Drive, San Diego, CA 92117 (
).
Photogrammetric Engineering & Remote Sensing
Vol. 84, No. 10, October 2018, pp. 619–627.
0099-1112/18/619–627
© 2018 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.84.10.619
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
October 2018
619
