Novel geolocation technology for geophysical sensors for detection and
discrimination of unexploded ordnance
Dorota A. Grejner-Brzezinska1, Charles Toth1,2, Hongxing Sun1 and Xiankun Wang1
Satellite Positioning and Inertial Navigation (SPIN) Laboratory1
Center for Mapping2
The Ohio State University
470 Hitchcock Hall, Neil Av., Columbus OH, 43221
[email protected]
Tel. 614-292-8787, Fax: 614-292-2957
Chris Rizos
School of Surveying & Spatial Information Systems
The University of New South Wales, Sydney, Australia
[email protected]
Tel: +61-2-93854205, Fax: +61-2-93137493
BIOGRAPHY
Dr. Dorota Brzezinska is a Professor and leader of the
Satellite Positioning and Inertial Navigation (SPIN)
Laboratory at The Ohio State University. She received her
MS and PhD in Geodetic Science from The Ohio State
University. Her research interests cover kinematic
positioning with GPS, precision orbit determination of
GPS/LEO, INS/GPS/image/LiDAR integration, mobile
mapping technology, and robust estimation techniques.
Since 2003, she has served at the ION Council at various
capacities, and is currently Eastern Region VicePresident; she is Vice-President of the International
Association of Geodesy (IAG) Commission 4,
Positioning and Applications and chair of IAG SubCommission 4.1, Multi-sensor Systems. She is the 2005
recipient of the US Geospatial Intelligence Foundation
(USGIF) Academic Research Award and the 2005 ION
Thurlow Award. She is a Fellow of the International
Association of Geodesy.
Dr. Charles Toth is a Senior Research Scientist at The
Ohio State University Center for Mapping. He received
an MS in Electrical Engineering and a PhD in Electrical
Engineering and Geo-Information Sciences from the
Technical University of Budapest, Hungary. His research
expertise covers broad areas of 2D/3D signal processing,
spatial information systems, high-resolution imaging,
surface extraction, modeling, integrating and calibrating
of multi-sensor systems, multi-sensor geospatial data
acquisition systems, and mobile mapping technology. He
is chair of ISPRS WG I/2 on LiDAR and InSAR Systems
and serves as the Director for the Photogrammetric
Application Division of ASPRS.
Dr. Hongxing Sun is a post-doctoral researcher in the
Satellite Positioning and Inertial Navigation (SPIN)
Laboratory, The Ohio State University. He received a
Bachelors degree in Geodesy, and MS and PhD degrees in
Photogrammetry, from Wuhan University, P.R. China.
His research interests include precise kinematic
positioning with GPS, GPS/INS integration for direct
platform orientation, and integrated multi-sensory
geospatial data acquisition systems.
Mr. Xiankun Wang is a PhD student and a Graduate
Research Associate in the Satellite Positioning and
Inertial Navigation (SPIN) Laboratory, The Ohio State
University. He received a Bachelor’s degree in Geodesy
from Wuhan University, P.R. China, and and MS degree
in Geoinformatics from the International Institute for
Geo-Information Science and Earth Observation (ITC),
The Netherlands. His research interests include GPS/INS
integration and terrestrial laser scanning applications.
Prof. Chris Rizos is a graduate of The University of New
South Wales (UNSW), Sydney, Australia; obtaining a
PhD in Satellite Geodesy in 1980. Chris is currently the
Head of the School of Surveying & Spatial Information
Systems at UNSW. Chris has been researching the
technology and applications of GPS since 1985, and
established over a decade ago the Satellite Navigation and
Positioning group at UNSW, today the largest and best
known academic GPS and wireless location technology
R&D laboratory in Australia. Chris is the Vice President
of the International Association of Geodesy (IAG), a
member of the Governing Board of the International
GNSS Service, and a member of the IAG's Global
Geodetic Observing System Steering Committee. Chris is
a Fellow of the IAG and of the Australian Institute of
Navigation.
ABSTRACT
Reliable and precise navigation technology is essential for
robust detection and discrimination of unexploded
ordnance (UXO) in a wide range of field conditions. The
detection and remediation of munitions and explosives-ofconcern (MEC) on ranges, munitions burning and open
detonation areas, and burial pits is one of the US
Department of Defense’s (DoD) most pressing
environmental problems. The MEC characterization and
remediation activities using currently available
technologies often yield unsatisfactory results, and are
extremely expensive, due mainly to the inability of
current technology to detect all MEC present at a site, and
the inability to discriminate between MEC and nonhazardous items due mainly to insufficient accuracy of
georeferencing of the geophysical images. As a result,
most of the costs (90%) of MEC site remediation are
currently spent on excavating targets that pose no threat.
Thus, the goal of the research presented here, supported
by a DoD’s Strategic Environmental Research and
Development Program (SERDP) grant, is to design and
demonstrate a high-accuracy hybrid navigation and
georeferencing device based on multi-sensor integration,
which can meet the stringent requirements of a manportable geophysical mapping system in open and
impeded environments, and hence to lower the cost of
remediation by improving the geolocation accuracy of
MEC discrimination.
This paper describes a hybrid system based on quadruple
integration of GPS, inertial technology (IMU), pseudolites
(PL) and terrestrial laser scanning (TLS) technology to
improve the current geolocation capabilities at MEC sites.
The concept design of the system, the algorithmic
approach to sensor integration, with a special emphasis on
TLS integration with GSP/IMU/PL, and the preliminary
performance assessment based on simulations are
presented.
1. INTRODUCTION: MOTIVATION
BACKGROUND INFORMATION
AND
Many Formerly Used Defense Sites (FUDS), such as
munitions burning and open detonation areas, and burial
pits, thousands of acres in size, are still contaminated with
unexploded ordnance (UXO). The detection and
remediation of MEC sites is one of the Department of
Defense’s (DoD) most pressing environmental problems.
Assessment tools that can efficiently survey large tracts of
land are required, and traditional ground-based survey
procedures are often too inefficient and costly. There are
aerial systems available, such as the Multi-sensor Towed
Array Detection System (MTADS) for large-scale UXO
geophysical surveys, which is an airborne version of a
similar
land-based
system
(http://www.nrl.navy.mil/pao/pressRelease.php?Y=2005
&R=17-05r). However, depending on the type and the
scope of the terrain, the types and amounts of buried
metal objects, airborne or land-based detection
technologies must be applied. The focus here is on landbased man-portable UXO detection system.
Geophysical mapping is the standard initial approach to
investigating MEC sites and identifying potential MEC
objects. In the standard practice, the first phase of
geophysics work is “mag and flag”, with the objective of
screening the entire site, and identifying some portions for
further investigation (local interrogation). “Mag and flag”
refers to the traditional way of finding UXO, where
technicians walk survey lanes (transects) laid out in a
field (see Figure 1), swinging a metal detector in front of
them. The most commonly used detector for UXO is a
magnetometer, hence the “mag”. However, this normally
can result in thousands of points being flagged as
potential MEC targets. Thus another step of local
interrogation is needed to determine the type of buried
object before the removal by excavation can be carried
out. Since excavation is the most expensive part of the
MES remediation process, it is naturally desirable to limit
the excavation to only those objects that pose a thread to
the public and the environment (compare an example
illustrated in Figure 2, which shows the difference
between the anomalies detected and identified as UXO).
The second phase normally involves an electromagnetic
(EM) survey of the site, to provide improved geophysical
data, usually along transects with increased density, over
the suspected target area. It is crucial that the images
acquired by the EM survey provide sufficient information
to detect all MEC present at a site and to discriminate
between MEC and non-hazardous items, as well to
identify a type of MEC/UXO. A complete process of
remediation of MEC/UXO sites is accomplished through
the following remedial activities:
•
MEC location and avoidance
•
MEC identification
•
MEC hazard assessment
•
Surface and subsurface MEC removal
•
MEC decontamination and destruction
•
Thermal and bioremediation of explosive
contaminated soil
= detected anomaly
Suspected target area
Site boundary
Hiking trail
Proposed
school location
Suspected firing point area
= Non-MEC anomaly
= MEC anomalies
(rockets and fragments)
= Practice Round
(60mm mortar w/
spotting charge)
Figure 1. Define boundaries of the target area and
geophysical transect spacing based on historical data,
such as, for example, Archive Search Report conducted
by the Army Corps of Engineers as part of the historical
records
review
process
(http://www.epa.gov/fedfac/documents/brief_dod_091205
.ppt)
MEC location, identification and hazard assessment are
crucial steps before the actual removal, and the process of
removal depends on several factors, and the type of
MEC/UXO is one of them. Thus, it is clear that
MEC/UXO identification, performed based on the
geophysical information, is central to the entire
remediation process, and the quality of EM imagery is the
key factor here.
The MEC hazard assessment determines the type of
remediation activity. Numerous factors or functional
relationships are considered in this assessment
(http://www.epa.gov/fedfac/documents/brief_dod_091205
.ppt):
• Severity: the potential severity of the result
should an MEC item function.
• Accessibility: the likelihood that a receptor will
be able to interact with an MEC item.
• Sensitivity: the likelihood that an MEC item will
function should a receptor interact with it.
Suspected
target area
Site boundary
Hiking trail
Proposed
school location
Suspected firing point area
Figure 2. Detected anomalies after “mag and flag” step
(top) and anomalies identified after refined
geophysical
mapping
(bottom)
(http://www.epa.gov/fedfac/documents/brief_do
d_091205.ppt).
The following are considered the major severity factors:
filler type, distance to additional potential human
receptors, proximity to critical infrastructure, proximity to
cultural resources, and proximity to ecological resources.
The input factors defining the accessibility are: site
accessibility, potential contact hours, amount of MEC,
and MEC depth relative to intrusive depth, migration
potential. The input factors that determine the level of
severity are: MEC category, and MEC size. All these
factors have a specified numerical score ranging from 115
to 1000, and are summed up to arrive at the final MEC
hazard
assessment
score
(see,
http://www.epa.gov/fedfac/documents/brief_dod_091205.
ppt for details). Clearly, the entire process requires
detailed information from historical sources and from
precise and accurate geophysical surveys, where
navigation, or EM sensor georegistration, are crucial, as
the information about the amount, location and type of
MEC/UXO is derived from the EM imagery.
1.1. Problem definition
As reported by SERDP (2005), based on actual field
experience, over 90 per cent of objects excavated during
MEC site remediation are found to be non-hazardous
items. In order to address this problem the Strategic
Environmental Research and Development Program
(SERDP)1 of DoD coordinates numerous efforts aimed at
developing new and improved technologies to
discriminate MEC from non-hazardous buried items.
According to SERDP, “using current sensor technologies,
the best hope for such discrimination lies in detailed
spatial mapping of magnetic or electromagnetic signatures.
Such investigation requires geolocation technologies that
function at two levels. First, anomalous signals must be
coarsely located so that they can be reacquired with a
required absolute accuracy of tens of centimetres. Second,
detailed mapping of signatures requires the measurement
of the locations of individual sensor readings to a relative
accuracy on the order of roughly 1 cm. By virtue of
topography or vegetation, many sites are not amenable to
Differential Global Positioning System (DGPS)” (SERDP,
2005).
No navigation system currently available can meet the
stringent navigation/positioning requirements as stated
above. One possible solution, as presented in this paper, is
to design a hybrid system based on several navigation and
imaging technologies that offer either good absolute
navigation accuracy or can support reliable relative
navigation in a local frame. The level of integration and
the capability of adapting the sensor and software
configuration to specific deployment scenarios are the key
characteristics of this approach.
The range of navigation techniques applicable to this task
includes satellite and terrestrial RF ranging systems,
inertial navigation systems (INS) based on inertial
measurement units (IMU), laser scanning systems, and
1
SERDP is the DoD’s environmental science and technology
program, planned and executed in full partnership with the
Department of Energy and the Environmental Protection
Agency, with participation by numerous other federal and nonfederal organizations. To address the highest priority issues
confronting the Army, Navy, Air Force, and Marines, SERDP
focuses on cross-service requirements and pursues highrisk/high-payoff solutions to the Department’s most intractable
environmental problems. The development and application of
innovative environmental technologies support the long-term
sustainability of DoD’s training and testing ranges as well as
significantly reduce current and future environmental liabilities
(http://www.serdp.org/)
even electro-optical devices such as total stations. Each
technology has its own limitations, which are expected to
be balanced within the integrated system design, where
complementary and often redundant characteristics of
these sensor technologies become advantageous. Thus,
the goal of the project described in this paper is to design,
implement and test a high-accuracy hybrid navigation
device that can address the stringent requirements of a
man-portable geophysical mapping system, and is capable
of maintaining high relative positioning accuracy in GPSchallenged environments.
2. SYSTEM DESIGN AND ARCHITECTURE
The system design, as illustrated in Figures 3 and 4, is
based on quadruple-integration of GPS (or more correctly
Differential GPS, or DGPS), inertial technology,
terrestrial RF system – pseudolite (PL), and terrestrial
laser scanning (TLS) that collectively can support highaccuracy geolocation of geophysical sensors in a range of
operating environments. The proposed design integrates
PL and GPS range signals together with the INS and TLS
measurements to deliver an optimal hybrid positioning
solution in a tight integration mode using the Extended
Kalman Filter (EKF) algorithms. Multi-sensor integration
is mandatory to assure accuracy, continuity and integrity
of the navigation solution. To facilitate high relative
positioning accuracy in a local frame, in GPS-challenged
environments, a novel integration approach is proposed
that incorporates TLS technology. To achieve an
environment-invariant performance of the TLS-based
positioning subsystem, easily deployable spherical ground
targets will be used.
System
GPS
PL
Advantage
Absolute location
Accurate
position/velocity
Absolute location
Works well in confined
environment (function
of
geometric
configuration)
Self-contained system
INS
TLS
Accurate position
Accurate attitude
Disadvantage
Line-of-sight
system
Line-of-sight
system
Relative position
Position
drifts
with time
Relative position
Table 1. Rationale for sensor integration, based on
complementary
characteristics
of
the
technologies.
Base
GPS Receiver
PL Transmitter
GPS Receiver
Rover
GPS Receiver
that will be suitable for specific deployment scenarios
taking into account constraints on weight and power,
operational time lines, and economic factors.
Base GPS
Rover GPS
PL antenna
Rover
PL Receiver
IMU Sensors
Tightly Coupled
GPS/INS/PL/TLS
Extended
Kalman Filter
Rover PL
Delta V
Delta
Position,
Velocity,
Attitude
Estimates
Strapdown
Navigation
Solution
GPS antenna
Top view of an UXO sensor pushcart
Laser scanner
GPS/INS/PL
laptop
Mounting pole
Laser scanner
laptop
UXO/MEC
detection
sensor
control
console
INS sensor
TLS System
Distance measured
Relative orientation
Figure 3. GPS/INS/PL/TLS integration: system design.
The three-tier positioning concept is hierarchical, with
several levels of positioning or sensor geolocation (see
also Figure 5):
1.
Absolute positioning in a global reference system
(WGS-84), applicable to open-sky areas, achieved
primarily using DGPS/INS integration.
2.
Relative medium-range positioning (with the
connection to global reference system preserved)
under canopies or other obstructions. The connection
to WGS-84 (or another selected mapping frame) is
accomplished be means of PL/INS technology, i.e.,
PL will substitute for GPS signals for medium to long
transmitter/receiver distances.
3.
Relative short-range positioning in a local frame
(with the connection to global reference system
preserved). This mode supports the detailed mapping
of geophysical signatures, where GPS is not available,
and where extremely high relative positioning
accuracy is required, which can be facilitated by TLS
technology supported by ground-based deployable
targets. Since the laser scanner reference frame is
connected to the GPS/INS/PL reference system via
lever arm and boresight angles, estimated by the
system calibration process, the absolute positioning
capability will be maintained. The accuracy of the
absolute coordinates may vary, and is a function of
the quality of the last acceptable GPS or PL-based
solution, while the relative positioning accuracy in
the local frame is expected to be at the few cm-level.
Currently a full-blown hardware and software prototype
configuration is being implemented, as shown in Figures
3 and 4. The future field tests and the expected software
and hardware tuning and optimization process will render
(or help define) the sub-categories of the implementation
Laptop (2)
Console
Cargo
area
to store
laser
targets
Geophysical
sensor
system
Side view of an UXO/MEC sensor pushcart
Figure 4. Sensor configuration for geophysical mapping
of MEC sites.
GPS satellites
Tethered balloon
with GPS receiver
and PL transmitter
PL transmitter
UXO sensor pushcart
Laser target
Relative medium-range positioning
Relative short-range positioning
Absolute positioning
Figure 5. Three-tier MEC site survey concept.
2.1. Navigation and imaging technologies used
With the maturity of processing algorithms and
availability of commercial GPS/INS integrated systems,
their use in civilian mapping applications as
georeferencing tools for imaging sensors is steadily
growing, and extensive literature on this subject exists
(e.g., Abdullah, 1997; El-Sheimy and Schwarz, 1999;
Grejner-Brzezinska and Toth, 1998; Grejner-Brzezinska,
1999; Grejner-Brzezinska, 2001a and b; GrejnerBrzezinska et al., 2005; Mostafa et al., 2000; Skaloud,
2002; Toth, 2002; Toth and Grejner-Brzezinska, 1998; Yi
et al., 2005). However, the navigation solution accuracy
provided by GPS/INS strongly depends on the quality and
geometry of GPS observations, as well as the type of IMU
used and the integration model applied. Generally, under
no or limited GPS availability, the navigation accuracy
that satisfies the requirements of sensor geolocation can
only be maintained for a few minutes if a navigationgrade IMU is used. Longer GPS gaps lead to a loss of
reliable geolocation information. Thus, alternative sensors,
such as PLs (e.g., LeMaster and Rock, 1999; Dai et al.,
2001; Grejner-Brzezinska and Yi, 2003; Barnes et al.,
2003a, b, 2005 and 2006), or imaging – optical or laserbased – systems (e.g., Veth and Raquet, 2006) must be
deployed to assure continuity of navigation solution.
• cm to dm-level positioning accuracy in static or
kinematic modes
• Single- and dual-frequency systems available
• Self-contained time-synchronized constellations
(such as Locata system used here) or asynchronous
PL networks.
The example commercial systems currently available are
the Terralite XPS technology from Novariant Inc.
(http://www.novariant.com/resources/technologies/positio
ning_infrastructure.cfm) and the LocataNet system from
Locata Inc. (http://www.locatacorp.com/).
2.1.1. Pseudolite technology
Pseudolites are ground-based radio navigation devices,
which are designed to support positioning and navigation
in situations where the GNSS (Global Satellite Navigation
System) constellation may be insufficient or unavailable
(e.g., indoor, urban or confined environments). PLs can
be configured to emit GPS-like signals for enhancing the
GPS by providing increased accuracy, integrity, and
availability, or to serve as independent navigation system
using non-GPS frequency signals. Accuracy improvement
due to PL signals, if combined with GPS, can occur
because of better local geometry, as measured by a lower
vertical dilution of precision (VDOP). Since the 1970s,
when the PLs were first used for testing the concept
design of the GPS system, their field of applications has
expanded considerably, even though their use is still
limited. The application markets that pseudolites are able
to serve range from construction monitoring, machine
control and open pit mining, and indoor and urban
navigation, through Location Based Services (LBS),
enhancement of the Space Based Augmentation System
(SBAS) supporting, for example, aircraft landing or
harbour entry and operations, to future applications in
Mars navigation (see, for example, Klein and Parkinson,
1986; Raquet et al., 1996; Elrod et al., 1996; LeMaster et
al., 2003; Montefusco et al., 2005; Progri et al., 2006;
Barnes et al., 2007).
The primary characteristics of the currently available PL
technology are:
• Line-of-sight (subject to signal blockage)
• Used as augmentation to GPS or an autonomous
constellation at
o GPS frequency, or
o Non-GPS frequency
• Typical signal range 3-5 km (depending on
transmitter power)
• mm to cm-level ranging accuracy achievable
Transmitting
antennae
Solar panels
Figure 6. Typical mount of a Locata transmitting antennae
with solar panels, Locata Test Facility,
Numarella, NSW.
The PL technology, used in this project, is based on
LocataNet (e.g., Barnes et al., 2003a and b, and 2005), a
network of dual-frequency self-synchronized groundbased transmitters (LocataLites) designed to cover a
survey area with ranging signals of continuous coverage
(Figures 6-7). These ranging signals transmit in the
license-free 2.4GHz Industry Scientific and Medical
(ISM) band. The LocataLite hardware design uses stateof-the-art field programmable gate array (FPGA) devices.
They provide configurable logic, on-chip memory and
digital signal processing (DSP) capabilities. For signal
spatial diversity, one LocataLite can be connected to two
transmit antennas, allowing for transmission of two
signals with different PRN codes at the same frequency.
A Locata receiver uses four or more ranging signals to
different LocataLites to compute a high-accuracy position
entirely independent of GPS. The Locata positioning
technology has been designed with four key objectives:
availability in most nvironments, high-reliability, highaccuracy, and cost effectiveness. Essentially, Locata
allows complete control over a ground-based PL
constellation, leading to an optimal positioning geometry
and consistent cm-level positioning accuracy. An
important feature of the Locata positioning signals is that
they are time-synchronized, which allows single-point
positioning similar to pseudorange-based GPS. However,
unlike GPS, the sub-cm level of synchronization between
LocataLites allows single-point positioning with GPSRTK (real time kinematic) level of accuracy without the
use of a reference station and data link. In addition,
Locata signal strength of up to 1 Watt is significantly
higher than the GPS signals, and thus offers better foliage
penetration within the range of a few to tens of kilometers.
(It should be noted that in stand-alone mode, Locata may
suffer from poor height accuracy if there is little variation
in elevation angle between the terrestrial transmitters and
receivers.) It should be mentioned that in operational
environements the coordinates of the LocataLite antennae
are normally surveyed using GPS and total station
equipment to provide WGS-84 Earth Centered Earth
Fixed (ECEF) coordinates.
Transmitting
antennae
Figure 7. Locata transmitting antennas mounted on 15 m
mast to provide sufficient vertical geometry for a
3D solution within a few tens of meters of the
mast (e.g., at ~30 from the mast, VDOP is ~4),
Locata Test Facility, Numarella, NSW.
The following are the operational steps required for
Locata and GPS systems to work together in real time
(Barnes et al., 2006):
•
GPS initialization – GPS RTK ambiguity resolution;
typically takes less than 1 minute with good satellite
geometry and tracking on at least 5 satellites.
•
Locata initialization – the GPS position and lever
arm offsets with respect to Locata receiving antenna
are input to the corresponding Locata receiver. The
Locata receiver applies the lever correction to convert
the input GPS position to the Locata receiver phase
center. This derived position is used to initialize the
Locata direct carrier range (DCR) positioning
algorithm.
•
GPS/Locata solution – once both GPS and Locata
solutions have been initialized, a cm-level position
solution is provided. Time synchronization between
Locata and GPS time systems is required for a
combined solution.
•
Locata-only solution – if at any time the GPS
solution fails to meet the accuracy specification due
to poor satellite geometry, satellite availability,
failure to resolve ambiguities or wireless data link
failure, the Locata receivers continue to deliver cmlevel positioning solution. Note that once the Locata
receiver position has been initialized, it operates
independently of GPS, and requires re-initialization
only if the number of LocataLite signals drops below
four (for 3D positioning). However, with dual
transmit LocataLites, providing signal spatial
diversity (see Figure 7), and careful LocataNet design,
the need for re-initialization is minimized.
2.1.2. Terrestrial laser scanning (TLS)
The ability of terrestrial laser scanner technology to
capture large amounts of spatial data quickly, accurately
and automatically makes it an important tool for spatial
data and surface geometry acquisition, even for complex
surfaces and environments often inaccessible by classical
survey techniques. TLS consists of an instrument for
measuring distance (laser) and a scanner. The focused
laser beam is directly reflected off the surveyed surface,
obviating the need to use prism reflectors. Generally 3D
environments are captured faster with TLS, as compared
to alternative techniques, with the accuracy ranging from
sub-millimeter on small object scans to 25 mm on objects
at
distances
up
to
250
m
(http://www.ceg.ncl.ac.uk/heritage3d/downloads/TSALaser-Scan.pdf). The output from a TLS system is a 3D
point cloud (X, Y, Z coordinates in a local TLS frame that
can be converted to a selected mapping frame, in which
the TLS location coordinates and orientation are defined).
TLS is normally placed at a site with known coordinates,
or GPS/INS can be used for direct geolocation if the
system is located on a moving platform. However, an
inverse problem can be defined, that is, if the TLS’s
location is unknown, it can be resected from the scanned
objects with known locations, or, if the absolute
coordinates of the scanned objects are not identified, the
objects’ coordinates defined in some local frame (or TLS
frame) can be used to determine a change in the TLS
location coordinates. This is possible by matching the
multi-site scans so as to determine the coordinate
transformation parameters between the sites. Thus, TLS
can be used as a navigation device to navigate either in a
global (absolute) reference frame or in a local frame, in a
relative mode. This application mode of TLS is utilized in
the hybrid navigation system discussed here. An added
benefit of using laser technology is that it is an active
sensor, and therefore there is no dependency on ambient
lighting conditions that may vary significantly in
vegetated areas.
The primary characteristics of the TLS technology are:
• Rapidly advancing surface measurement technology
• Existing implementations
o Pulsed scanner (Time-of-Flight, TOF)
o Continuous waveform scanner (Phase)
o Flash LADAR2 or 3D camera
• Primary specifications
o Large number of range measurements (5-200K
points/s)
o mm-level ranging accuracy
o Dense spatial sampling (0.01° angular
sampling rate)
o Excellent point cloud relative accuracy (3D)
o Sensitive to moving objects (scanning)
• Large variety of analytical surface modeling
techniques (scan data analysis software)
• Surface/patch matching technique can be used to
detect change in TLS location that can be used to
support navigation
o Similarity
transformation
(7-parameter)
between two surfaces at locations T1 and T2 in
Figure 7
o Feedback to navigation system
Examples of the currently available TLS technology are
Cyrax 2500 from Cyra Technologies/Leica Geosystems,
Optech's ILRIS Intelligent Laser Ranging and Imaging
Systems, and Trimble GX terrestrial laser scanner used in
the system described here. A broad list of TLS technology
references can be found at http://www.itc.nl/isprswgiii3/resources_links.html#TerrestrialScanners.
Cyrax 2500, considered one of the most commonly used
terrestrial scanning systems, offers a single-point range
accuracy of ± 4 mm, angular accuracies of ± 60 μrad, and
a beam spot size of only 6 mm over a 0-50 m range. Its
360° by 195° pan and tilt mount and dual internal rotating
mirrors enable it to be deployed in virtually any
2
Laser Distance Array (LADAR)
orientation
(http://www.meabmx.se/tidning/cyrax2500.htm). The point density is
dependent on the object distance, with a 100 pts/m2
density typical at about 20 m ranges.
Optech’s ILRIS includes an on-board high resolution
digital camera. It offers scanning ranges from 3 m to
<1,500 m with the range accuracy of 3 mm and spot
spacing of 2.6 mm at 100 m range distance, and data
sampling
rate
of
2,000
points/s
(http://www.optech.ca/i3dhome.htm).
Trimble’s GX terrestrial laser scanner offers a standard
range of up to 200 m and in the ‘overscan’ mode, up to
350 m. The field of view is limited by 360° × 60°, with an
asymmetrical vertical part of about 40° above the horizon.
The scanning speed is up to 5,000 points per second. The
scanner uses an auto focus technique, which guarantees a
constant small laser spot even at different distances within
a scan. The size varies from 0.3 mm at 5 m, up to 1.5 mm
at a distance of 25 m.
xL
T1
xL
xL
T2
yL
yL
yL
zL
zL
zL
E
A
D
B
C
Figure 7. TLS-based navigation concept based on
spherical target surface matching at two
locations of the TLS, T1 and T2. Targets are
denoted as A-E, the reference system XLYLZL is
centered at the TLS system and changes its
location and orientation between consecutive
locations of TLS.
2.2. System design and implementation
The system architecture, illustrated in Figure 11,
represents an extension and re-design of the original OSU
GPS/IMU integrated system, AIMS™ (e.g., GrejnerBrzezinska and Toth, 1998; Grejner-Brzezinska, 1999).
The novel quadruple integration of GPS (GNSS), PL and
inertial technologies, augmented by TLS, replace the tight
GPS/IMU integration module of AIMS™. The following
are the primary characteristics of the new AIMS-PRO™
system and software design:
•
•
•
•
•
•
•
•
Handles various IMU classes, from navigationthrough tactical to consumer-grade sensors
Loose and tight integration modes (tight mode
illustrated in Figure 11)
o Ultra-tight considered (currently, optional)
Implements feedback from imaging module to
implement terrain-referenced navigation in GPSchallenged environments
o TLS with specialized targets for highest
navigation accuracy
o Airborne laser scanning (ALS)
o Multiple image overlap based relative
orientation from optical imagery, DEM/GIS,
etc.
Multiple approaches to GPS solution (in loose
integration mode)
o Single baseline differential kinematic
solution
o Network-based
differential
kinematic
solution
o Precise point positioning technique (PPP)
Post-processed and real-time capable
Improved portability (less platform-dependence)
Flexible design
Improved error diagnostics capabilities
Figure 8. Screenshot of AIMS-PRO™ interface: data
input.
Figure 9. Snapshot of AIMS-PRO interface: KF
parameter setting.
Figures 8-10 are examples of the AIMS-PRO™ interface
design. The input interface is designed for handling a
variety of input data types, including initial location and
orientation parameters (Figure 8). The data processing
interface handles all configuration and control parameters
defining the data processing modes. For example, the
initial parameter settings for EKF allow for selection of
the sensor models suitable for a particular class of sensors
or a specific IMU (Figure 9). GNSS processing
parameters, error tolerances, ambiguity resolution
parameters, etc., are also defined through this interface
(Figure 10).
Figure 10. Screenshot of AIMS-PRO™ interface:
configuration and control parameters.
GPS raw measurement
Locata carrier phase pseudorange
IMU angular rate and specific
force
Ambiguity resolution
Initial carrier phase bias resolution
INS navigation
GPS double-differenced carrier
phase and/or pseudorange
Carrier phase range
Position, velocity, and attitude
Kalman filtering
INS error estimates
ALS/TLS measurements
Imaging system
Feedback from observed
position and attitude data
Spherical target/DEM extraction
Feature extraction
Spherical target parameters
Spherical target/DEM matching
Feature parameters
Feature matching
Combining TLS and image info
Figure 11. AIMS-PRO™ design architecture.
2.3. TLS-based navigation: the concept
In the concept of a multi-sensor geolocation system
presented here, TLS is used to support navigation in
wooded areas, where GPS may not be available, and the
PL network may be subject to increased multipath and
partial signal blockage. High-accuracy reference surfaces
measured by TLS can be matched to sub-cm accuracy to
detect relative motion of the platform carrying the
MEC/UXO (or EM) sensor assembly. Using a rigorous
least squares 3D surface matching, complex surfaces can
be matched at the level of the ranging accuracy (see, for
example, Gruen and Akca, 2005). The accuracy of
matching is measured in terms of the accuracy of the
relative orientation parameter estimates, including
position and attitude data (3+3). To assure good quality
matching results, multiple spherical targets are used
within the survey area. The targets are portable, easily
deployable on vertical poles or placed on the ground, and
provide surface control to connect the scans performed at
different platform positions. The range determined
between the center of the target and TLS is used to find
the coordinates of TLS by resection or partial resection
(in 2D), as shown in Figure 3. The spherical targets can
be described with four parameters: three coordinates of
the center of the sphere and its radius (known); thus, in
theory, a sphere can be determined from three laser points
reflected from its surface. Considering the random
ranging errors, more points are needed to assure both the
robustness of the estimation and the accuracy of surface
matching, and estimation of the centers of the spheres by
the least squares adjustment method.
2.3.1. Determination of the spherical target coordinates
Determination of the spherical target centers from TLS
point cloud includes the following three steps.
1.
Look up table (LUT) indexing: indexing TLS point
cloud with a Look Up Table (LUT) is a simple, yet
efficient method for TLS data processing. The basic
idea of LUT indexing is establishing a link between
laser points and a two-dimensional array through
azimuth and vertical angles of the points. Thus, a
point can be directly accessed through the memory
address stored in LUT without having to search the
whole dataset. More importantly, LUT arranges the
randomly distributed data into order by indexing
points in space with the neighboring cells in the table;
clearly, it benefits the sphere point classification
algorithm.
2.
Sphere point classification: two algorithms have
been developed for sphere point classification, (1)
shortest-range algorithm, and (2) region growing
algorithm. The shortest-range algorithm is based on
finding the shortest range (distance) from the scanner
to the surface of the spherical target within a search
window, and selecting points with ranges within the
threshold
defining
the
shortest
range
( Δs ≤ s − 2rs − s , where s is the range and r is the
sphere’s radius). Next, the least squares fitting
algorithm is applied to determine the approximate
center candidate from the selected points on the
sphere, as described in the next section. The region
growing algorithm is based on segmentation of the
points within the search window into different groups
(objects), and subsequently performing the fitting of
the points from each group onto the sphere. Based on
the simulations performed to date, the region growing
algorithm was found to perform slightly better,
particularly in case of occlusions and noisy data. See
Grejner-Brzezinska et al. (2008) for the results of
using shortest-range algorithm for spherical target
center determination.
2
Both sphere point classification algorithms are
followed by a sphere center refinement procedure to
assure inclusion of the maximum number of genuine
sphere points into the sphere center estimation
process. Since there is a certain overlap between
search windows, two neighboring windows which
cover the same sphere region will generate two
approximate sphere centers. During the refinement
step, the distances between the sphere center
candidates are checked against a predefined threshold
(0.05m in tests presented here); if the condition is
met, the coordinates from these candidate solutions
are averaged. Moreover, after finding the shortest
range, s, by means of the averaged sphere center
coordinates, the search window width, D, can be
D=
2r s 2 + 2rs
s+r
determined as
, to search the entire
range of sphere points, to assure that the maximum
number of sphere points is properly discriminated.
the approximate sphere center coordinates can be
computed. Generally, the fitting process converges within
a few iterations with data having range noise of up to 2
cm.
2.3.2. TLS-based resection problem
In the actual application, if a geophysical signal is
detected during the traversing of an MEC site, the highresolution/accuracy local survey may be needed for cued
interrogation. In this case 6-10 spherical targets are placed
around the border of a local area of about 10-20 m by 1020 m (Figure 12). A near uniform distribution of the
targets is desirable, but for operational purposes,
depending on the terrain structure and the surrounding
environment, some flexibility of the target distribution is
allowed (the geometric configuration as well as the
vertical distribution of these targets is currently the
subject of ongoing simulation studies). It is important to
note that there is no need to position the targets, and the
only requirement is that they should not be moved during
the local survey.
x p = xC + s ⋅ cos β sin α
y p = yC + s ⋅ cos β cos α
(2)
z p = z C + s ⋅ sin β
The TLS range measurements are best described in a
polar coordinate system centered at TLS, and the
coordinates of a measured point in the mapping frame can
be calculated using Eq. (2), where x p , y p , z p and
(
)
( xC , yC , zC ) are the vectors of coordinates of a measured
point and the laser scanner, respectively, in the mapping
frame, s is the distance from the laser scanner to the target,
α is the azimuth, and β denotes the vertical angle.
D
3.
E
Least squares fitting: assume that the approximate
sphere center is ( x0, y0 , z0 ) the ith measured sphere
point is ( xi , yi , zi ) , r is the sphere radius (known and
3
C
constant). Then:
2
( x 0 − x i )2 + ( y 0 − y i )2 + ( z 0 − z i )2 = r 2
r
x2,1
F
(1)
After linearization, the residual of the ith “observed”
radius can be denoted as:
1
B
r
x A,2
r
xA,1
Spherical targets
Scanning Sites
Scanning lines
Vri =
(x0 − xi )
ri
δx +
( y0 − yi )
ri
δy +
(z 0 − z i )
ri
δz + r − ri
(2)
where (δx, δy, δz ) denote coordinate corrections to the
approximated sphere center coordinates, and ri denotes
the radius to the approximate sphere center. By
employing the least squares adjustment, the corrections to
A
Track System’s path
Figure 12. The spatial (3D) or partial resection (2D)
concept using TLS ranges and spherical targets
(Grejner-Brzezinska et al., 2008).
As explained above, the coordinates of the centers of the
spherical targets (points A-F in Figure 12), relative to the
TLS coordinate system, can be calculated from the
coordinates of the point cloud on the target’s surface. To
simplify the computations, a local coordinate system is
introduced, with the origin at site 1 (Figure 12). The
spatial relationship between the two consecutive TLS
locations, 1 and 2, with respect to target A, can be
described by a rigid body transformation, including three
offsets and three rotation angles (Eq. (3):
r
r b1
r
x Ab1,1 = x2,1
+ Rbb21 x Ab ,22
(3)
rb2
where x A,2 is the translation vector from site 2 to point A
r b1
in the local TLS coordinate system of site 2, x2,1 is the
translation vector from site 1 to site 2 in the local TLS
r b1
coordinate system of site 1, x A,1 is the coordinate vector
from point 1 to site A in the local TLS coordinate system
of site 1,
Rbb21
is the rotation matrix from the local
TLS coordinate system of site 2 to that of site 1. For
multiple points, Eq. (3) can be expressed in a matrix form:
b1
X Pb1,1 = X 2,1
+ Rbb21 X Pb 2,2
(4)
where P denotes all common targets measured at sites 1
and 2. In Eq. (4) there are six unknown parameters: the
b1
three translation parameters in X 2,1 , and three Euler
angle parameters in Rbb21 . To derive a more convenient
expression, Eq. (4) is multiplied by the rotation matrix,
Rbn1 , from TLS coordinate system of site 1 to the
navigation coordinate system; then, the coordinate
transformation, expressed in the navigation coordinate
system, is given by:
n
Rbn1 X Pb1,1 = X 2,1
+ Rbn2 X Pb 2,2
(5)
Adding the coordinates of site 1 in the navigation frame to
n
,
both sides of Eq. (5), considering that X 2n = X 1n + X 2,1
gives:
X 1n + Rbn1 X Pb1,1 = X 2n + Rbn2 X Pb 2,2
(6)
Eq. (6) can be linearized assuming small angular
differences in the rotation matrix:
X 1n + Rbn1 X Pb1,1 = X 2n 0 − δ X 2n + ( I − E ) Rbn2 0 X Pb 2,2
(7)
where I is the identity matrix, E is the skew-symmetric
matrix of the attitude angle error, and δ X 2n is the
coordinate error vector. Rearranging Eq. (7) provides the
final positioning and attitude determination equation that
is fed directly to the Kalman filter (Eq. 8). Note that it
contains information that can be used to calibrate IMU
errors.
X 1n + Rbn1 X Pb1,1 − X 2n 0 − Rbn2 0 X Pb 2,2 = −δ X 2n + X Pn ,2 0ε
where
ε
is the attitude angle error vector, and X
(8)
n 0
P ,2
is
the skew-symmetric matrix of the coordinate vector. Note
that the relative accuracy provided by Eq. (8) might be
high, but the resulting navigation accuracy depends on the
accuracy of the position and attitude of site 1, which is
assumed to be determined by the integrated
DGPS/PL/INS.
3. TLS-BASED NAVIGATION: SIMULATIONS
TLS point cloud data was simulated and a simple
algorithm for the determination of the center of a
spherical target was developed. The point cloud data set
was generated based on the actual scanning pattern of a
terrestrial laser scanner. One of the simulated scenarios is
a flat ground with an area of 4 m × 4 m, with one side
surrounded by a 3-meter high wall, and two spherical
targets, with radii of 0.20 m, placed on the ground with
center coordinates of (1.0, 3.5, 0.2) and (3.0, 3.0, 0.2), as
shown in Figure 13. The effect of two columns, located
0.15 m and 0.20 m, in front of the spheres were also
simulated in this scenario, so that most of sphere 1
centered at (1.0, 3.5, 0.2) is occluded, and sphere 2
centered at (3.0, 3.0, 0.2) is partially occluded by both
columns (Figure 13). The laser scanner was located at
(0.5, 0.5, 2.0); the assumed vertical angle for scanning
ranges was within ±75° and scan angle resolution of 0.5°
was assumed in both vertical and horizontal directions.
There were 33 data points simulated on sphere 1, and 42
data points on sphere 2. A random noise of 1 cm has been
added to each coordinate component. The region growing
algorithm, using a 16×16 (that is 16 times the scanning
angle resolution) search window and 9 cm distance
threshold for the neighboring cells, was used, followed by
the least squares fitting, to determine the coordinates of
the sphere centers from the distorted point cloud data. The
results are discussed below.
The coarse (approximated) and the final center
coordinates of the spheres were estimated, as shown in
Table 2. For the coarse result, the center coordinates of
sphere 1 were determined twice due to the 50% overlap
between the neighboring search windows. In the case of
sphere 2, only points on the right side of the sphere
contributed to the estimation of its center, since the left
side was segmented to different point groups, with a
fitting error of 0.011 (that is 10% higher than the selected
tolerance level for the fitting algorithm, which was equal
to 1 cm), and thus, it was not selected as the center
candidate. Apart from the correct candidates, there are
four false sphere centers fitted from the segmented points.
It was found that all the false sphere centers were
generated from points at locations with extreme vertical
angles. After the refinement procedure, these false center
candidates were removed and the accuracy of sphere 2
was improved, but the position accuracy of sphere 1
remained unchanged. Note that the success rate is defined
as percent of points properly identified on the sphere,
based on the comparison of the known reference
coordinates of each simulated target and the coordinates
determined by the algorithm described here.
Occluding
columns
Points on partially
occluded spherical
targets
implemented in the tightly-coupled mode under the EKF
architecture, PL module under design and implementation,
and the TLS module implemented and undergoing
extensive simulation testing. The preliminary results of
the estimation of the coordinates of the spherical target
center, crucial to the concept of TLS-based navigation,
proved to be of good accuracy, even for a relatively high
noise of 2 cm in the simulated point cloud coordinates,
and under partial target occlusions. Obviously, the
success rate of identifying the sphere points decreases
with the growing data noise and percentage of occlusions.
More tests and simulations are currently under way.
Acknowledgement: This research is supported at OSU
by the 2007 SERDP grant. The research into Locata
technology at UNSW is supported by an Australian
Research Council Discovery grant.
REFERENCES
1.
TLS location
Figure 13. Simulated TLS data set with occlusions and
1cm noise within the region covered by one
search window.
Another data set was simulated with eight spherical
targets each with radius 0.20 m, distributed on flat ground
over an area of 30×30 m2, and distances to the laser
scanner ranging from 2.7 m to 28.3 m. The estimated
position coordinates of the target spheres and the laser
scanner are listed in Table 3. With the 16×16 search
window and the distance threshold of 12 cm, and the
minimum point number for the least squares fitting set at
six, all spheres were extracted with the region growing
algorithm, as shown in Table 4. It was determined that all
spheres could be uniquely determined with the distance
threshold ranging from 9 cm to 20 cm. Similar results
were found using the shortest-range algorithm. Next, 2 cm
noise was added to all target center coordinates, and the
process was repeated, with the distance threshold of 12
cm and the minimum point number for the least squares
fitting of 14. The results are listed in Table 5. Notice that
sphere 2 is missing in this case. When the minimum point
number for sphere fitting was lowered to seven or less,
sphere 2 could be extracted, but some false sphere centers
were also derived from the ground points.
2.
3.
4.
5.
6.
4. CONCLUSIONS
This paper describes a concept design of quadruple
integration of GPS/INS/PL and TLS, used here as a
navigation-supporting sensor. The system is currently
under implementation, with GPS/INS modules already
7.
Abdullah, Q., (1997). Evaluation of GPS-Inertial
Navigation System for airborne photogrammetry,
presented at 1997 ACSM/ASPRS Annual Convention
and Exposition, 7-10 April, Seattle, WA.
Barnes, J., Rizos, C., Wang, J., Small, D., Voigt, G.,
& Gambale, N. (2003a). Locata: A new positioning
technology for high precision indoor and outdoor
positioning, 16th Int. Tech. Meeting of the Sat Div of
the U.S. Inst. of Navigation, Portland, OR, 9-12
September, 1119-1128.
Barnes, J., Rizos, C., Wang, J., Small, D., Voigt, G.,
& Gambale, N. (2003b). High precision indoor and
outdoor positioning using LocataNet, 2003 Int. Symp.
on GPS/GNSS, Tokyo, Japan, 15-18 November, 9-18,
CD ROM.
Barnes, J., Rizos, C., Kanli, M., Pahwa, A., Small, D.,
Voigt, G., Gambale, N., & Lamance, J., (2005). High
accuracy positioning using Locata’s next generation
technology, 18th Int. Tech. Meeting of the Sat. Div.
of the Institute of Navigation, Long Beach, CA, 1316 Sept, 2049-2056.
Barnes, J., Rizos, C., Kanli, M., Pahwa, A. (2006). A
solution to tough GNSS land applications using
terrestrial-based
transceivers
(LocataLites),
Symposium on GPS/GNSS (IGNSS2006), Surfers
Paradise, Australia, 17-21 July, CD-ROM.
Barnes, J., Lamance, J., Lilly, B., Rogers, I., Nix, M.,
Balls, A. (2007). An integrated Locata & Leica
Geosystems positioning system for open-cut mining
applications, Proceedings, 20th Int. Tech. Meeting of
the Sat. Div. of the Institute of Navigation,
September 25-28, Fort Worth, TX, 1661-1668.
Dai, L., Rizos, C., Wang, J. (2001). The role of
pseudo-satellite signals in precise GPS-based
positioning, Journal of Geospatial Eng., HK Inst. of
Engineering Surveyors, 3(1), 33-44.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
Elrod, B., Barltrop, K., Stafford, J., Brown, D. (1996).
Test results of local area augmentation with an inband pseudolite, Proceedings, ION NTM, Santa
Monica, CA, 22-24 January, 145-154.
El-Sheimy, N., Schwarz, K.P., (1999). Navigating
urban areas by VISAT – A Mobile Mapping System
integrating GPS/INS/Digital Cameras for GIS
application, Navigation, 45(4), 275-286.
Grejner-Brzezinska, D.A., Da, R., Toth C., (1998).
GPS error modeling and OTF ambiguity resolution
for high-accuracy GPS/INS integrated system, J. of
Geodesy, 72(11), 628-638.
Grejner-Brzezinska, D.A. (1999). Direct exterior
orientation of airborne imagery with GPS/INS
system: Performance analysis, Navigation, 46(4),
261-270.
Grejner-Brzezinska, D.A. (2001a). Mobile Mapping
Technology: Ten years later, Part I, Surveying and
Land Information Systems, 61(2), 79-94.
Grejner-Brzezinska, D.A. (2001b). Mobile Mapping
Technology: Ten years later, Part II, Surveying and
Land Information Systems, 61(3), 83-100.
Grejner-Brzezinska,
D.A.,
Yi,
Y.
(2003).
Experimental GPS/INS/Pseudolite system for
kinematic positioning, Survey Review, 37(288), 113126.
Grejner-Brzezinska, D.A., Toth, C., Yi, Y. (2005).
On improving navigation accuracy of GPS/INS
systems, Photogrammetirc Engineering & Remote
Sensing, April 2005, 71(4), 377-389.
Grejner-Brzezinska, D.A., Toth, C., Sun, H., Wang,
X., Rizos, C. (2008). GPS/INS/PL/TLS integration
supporting navigation of geophysical sensors for
unexploded ordnance detection and discrimination,
13th FIG International Symposium on Deformation
Measurements and Analysis, 4th IAG Symposium on
Geodesy for Geotechnical and Structural Engineering,
12-15 May, Lisbon, Portugal, CD ROM.
Gruen, A., Akca, D. (2005). Least squares 3D surface
matching, ASPRS 2005 Annual Conference,
Baltimore, MD, 7-11 March, CD ROM.
Klein, D., Parkinson, B.W. (1986). The use of
pseudo-satellites for improving GPS performance,
Global Positioning System, Vol. III, Institute of
Navigation, Washington, DC.
LeMaster, E.A., Matsuoka, M.,. Rock, S.M. (2003).
Mars navigation system utilizes GPS, IEEE
Aerospace and Electronics Systems Magazine, 18(4),
April 2003, 3-8.
LeMaster, E.A., Rock, S.M. (1999). Mars exploration
using self-calibrating pseudolite arrays. ION GPS1999, Nashville, TN, 14-17 September, 1549-1558.
Montefusco, C., Ventura-Traveset, J., LucasRodriguez, R., Toran, F., Jofre, M. (2005).
Enhancing SBAS performance: the EGNOS
pseudolite
concept
(http://www.egnospro.esa.int/Publications/EGNOS%20pseudolite%20E
22.
23.
24.
25.
26.
27.
28.
29.
30.
NC%202005_final.pdf, last accessed on April 3,
2008).
Mostafa, M., Hutton, J., Lithopoulos, E. (2000).
Ground accuracy from directly georeferenced
imagery, GIM International, 14(12), 38-41.
SERDP, Strategic Environmental Research and
Development Program (2005). SON NUMBER:
MMSON-07-01, 10 November 2005.
Progri, I.F., Maynard, J., Michalson, W.R., Wang, J.
(2006). The performance and simulation of a
CCDMA pseudolite indoor geolocation system, 19th
Int. Tech. Meeting of the Sat. Div. of the Institute of
Navigation, Fort Worth, TX, 26-29 September, 24652477.
Raquet, J., Lachapelle, G., Qiu,, W., Pelletier, C.,
Nash, T., Fenton, P., Holden, T. (1995).
Development and testing of a mobile pseudolite
concept for precise positioning, 8th Int. Tech.
Meeting of the Sat. Div. of the Institute of Navigation,
Palm Springs, CA, 12-15 September, 817-826.
Skaloud, J., (2002). Direct georeferencing in aerial
photogrammetric
mapping,
Photogrammetric
Engineering & Remote Sensing, March 2002, 207210.
Toth, C., Grejner-Brzezinska, D.A. (1998).
Performance analysis of the Airborne Integrated
Mapping System (AIMS™), ISPRS Commission II
Symposium on Data Integration: Systems and
Techniques,
International
Archives
of
Photogrammetry and Remote Sensing, Vol. XXXII,
Part 2, 320-326.
Toth, C. (2002). Sensor integration in airborne
mapping, IEEE Transactions on Instrumentation and
Measurement, 51(6), 1367-1373.
Veth, M., Raquet, J. (2006). Fusion of low-cost
imaging and inertial sensors for navigation, 19th Int.
Tech. Meeting of the Sat. Div. of the Institute of
Navigation, Fort Worth, TX, 26-29 September, 10931103.
Yi, Y., Grejner-Brzezinska, D.A., Toth, C. (2005).
Performance analysis of a low cost MEMS IMU and
GPS integration, ION Annual Meeting, 26-29 June,
1026-1036.
Coarse
Result
Final
result
True
value
x
(m)
y
(m)
z
(m)
Pos. Err
(mm)
σ̂ 0
0.999
0.997
3.003
2.695
1.345
0.031
0.042
0.997
3.003
1.0
3.0
3.501
3.503
3.000
2.732
0.050
0.034
0.962
3.503
3.000
3.5
3.0
0.195
0.193
0.201
2.682
-0.194
0.198
0.198
0.193
0.199
0.2
0.2
4.9
8.0
3.4
0.009
0.009
0.004
0.008
0.010
0.008
0.009
0.009
0.008
8.0
3.5
Estimated points
Sphere 1 Sphere 2
21
22
0
0
0
0
0
22
0
33
Other
0
0
16
0
0
0
0
0
33
0
0
0
15
41
33
45
0
0
Success
Rate
63.6%
66.7%
38.1%
false
false
false
false
66.7%
71.4%
42
Table 2. Comparison between the estimated sphere centers and the reference sphere center coordinates; simulated data with
1 cm noise.
Laser Scanner
Sphere 1
Sphere 2
Sphere 3
Sphere 4
Sphere 5
Sphere 6
Sphere 7
Sphere 8
x
(m)
y
(m)
z
(m)
Distance to laser
scanner (m)
Points
8.0
8.0
28.0
26.0
26.0
10.0
10.0
2.0
2.0
8.0
25.0
28.0
12.0
4.0
10.0
8.0
2.0
16.0
2.0
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
-17.1
28.3
18.5
18.5
3.4
2.7
8.7
10.2
-44
8
38
33
1365
2438
179
135
Table 3. The simulated data characteristics.
Sphere
x
(m)
y
(m)
z
(m)
Pos. Err
(mm)
σ̂ 0
1
2
3
4
5
6
7
8
8.000
28.000
26.000
26.000
10.000
10.000
2.000
2.000
25.000
28.000
12.000
4.000
10.000
8.000
2.000
16.000
0.2000
0.200
0.200
0. 200
0. 200
0. 200
0. 200
0. 200
0.03
0.05
0.01
0.01
0.00
0.00
0.01
0.00
0.00003
0.00003
0.00003
0.00003
0.00003
0.00003
0.00003
0.00003
Estimated points
Success
Sphere Other
Rate
44
0
100%
8
0
100%
38
0
100%
33
0
100%
1328
0
97.3%
2181
0
89.5%
179
0
100%
135
0
100%
Table 4. Derived sphere centers, simulated data with no noise.
Sphere
x
(m)
y
(m)
z
(m)
Pos. Err
(mm)
1
3
4
5
6
7
8
7.981
26.002
25.999
10.004
9.975
1.998
2.000
24.993
11.991
4.012
9.978
8.002
2.009
16.001
0.191
0.200
0.210
0.199
0.204
0.205
0.201
21.9
8.9
16.3
21.8
25.5
10.1
1.7
Estimated points
σ̂ 0
Points
in LUT
Sphere
Other
Success Rate
0.017
0.017
0.017
0.015
0.013
0.017
0.015
34
30
27
914
1677
130
95
28
30
27
644
719
121
91
0
0
0
0
0
0
0
63.6%
79.0%
81.8%
47.2%
29.5%
67.6%
67.4%
Table 5. Derived sphere centers, simulated data with 2 cm noise