UPDATING DIGITAL GEOGRAPHIC DATABASE USING VEHICLE-BORNE LASER
SCANNERS AND LINE CAMERAS
Huijing Zhao, Ryosuke Shibasaki
Center for Spatial Information Science, Univ. of Toky
{chou, shiba}@skl.iis.u-tokyo.ac.jp
Commission III, WG 6
KEY WORDS: Data Fusion, Vehicle-borne, Laser Range Scanner, Urban 3D
ABSTRACT:
VLMS is a mobile mapping system, where three single-row laser range scanners, six line CCD cameras as well as a GPS/INS based
navigation unit are mounted on a van, measuring object geometry as well as texture along the street. This paper contributes to a
method of fusing the data output of VLMS with existing geographic data sources, where focus is cast on the rectification of
GPS/INS parameters, which might be quite erroneous in urban area. An algorithm is developed to correct four parameters of each
GPS/INS update, i.e. xyz coordinates of vehicle position and yaw angle of vehicle orientation, by registering the laser points of
VLMS with an existing data source, e.g. DSM. The algorithm is examined using the VLMS data that are taken in GINZA area,
Tokyo. Manually assigning 18 sets of tie-points, GPS/INS parameters are corrected automatically and efficiently, so that the laser
points of VLMS are matched to a DSM. In data fusion, a set of objects are extracted from the rectified VLMS data using an interface
that was developed in our previous research, which contains of commercial sign board, traffic sign/signal, road boundary, road lights
and so on. They are integrated with a 1:2500 3D map that consists of building frames only. In addition, line images of VLMS are
projected onto the building facades of the 3D map, and textures are generated in an automated way.
1. INTRODUCTION
Up to now, many research efforts in photogrammetry and
remote sensing community have been devoted to the study of
aerial or satellite based mapping technologies for the
reconstruction of 3D urban objects (e.g. Collins 1994, Gruen
1998), and a vast amount of geographic database has been
established. Normally, aerial or satellite based survey can cover
relatively wide area, but fail in capturing urban details due to
the limitation of spatial resolution and viewing angle. On the
other hand, mobile mapping system (MMS) has emerged
recently as a promising survey technique for collecting detailed
spatial data from the ground. With the development of
automobile navigation system, 3D GIS (Geographic
Information System), ITS (Intelligent Transportation System),
and applications using virtual and augmented reality, 3D urban
database containing the details, such as sidewall (facade) of
buildings, traffic sign/signal, commercial signboards etc., are
found of increasing importance. A number of mobile mapping
systems have been developed (e.g. Ellum and El-Sheimy 2000,
He and Orvets 2000, Silva et.al.2000). A comprehensive
examination of MMS can be found in Li 1997 or El-Sheimy
1999. The mapping technologies from either air or ground have
their advantages and drawbacks. It is demonstrated that MMS
has efficiency in generating detailed spatial database. Whereas,
it lacks spatial coverage, which can be compensated using other
traditional survey techniques taking data from the air or satellite.
It is important that the data Sources collected using both
technologies are fused, so that a more comprehensive
geographic database can be generated.
A mobile mapping system called VLMS (Vehicle-borne
Laser Measurement System) has been developed in a joint
research effort between Asia Air Survey Co. Ltd and ours,
aiming at collecting the detailed spatial data in central urban
area. Except the GPS (Global Positioning System)/INS (Inertial
Navigation System) based navigation unit, three single-row
laser range scanners (briefly called "laser scanner") and six line
CCD cameras (briefly called "line camera") are mounted on a
van, mapping object's geometry as well as texture along streets.
A framework for automatically reconstructing textured 3D
models of buildings, roads and trees using vehicle-borne laser
range and line images (Zhao and Shibasaki 2003a), and an
interface for extracting a broad range of urban objects, such as
commercial signboard, road boundary, traffic sign/signal,
telegram pole/cable and so on, in a semi-automatic manner
(Zhao and Shibasaki, 2003b) were developed. Efficiency of the
system for generating a database of urban details was
demonstrated through a number of real experiments in central
Tokyo. However, since both laser range and line images are
geo-referenced directly using the position and orientation
parameters from GPS/INS based navigation unit, the geo-
referenced data sets as well as the modeling results might be
quite erroneous especially in central town, and have to be
rectified before being exploited to update existing geographic
database.
In the GPS/INS based navigation unit, GPS measures
vehicle's position using satellite signals, which might be heavily
obstructed by bridges, trees, tunnels and buildings in urban area.
INS, consisting of accelerometers and gyroscopes, measures the
velocity and direction changes of the vehicle with high
accuracy for only short periods. Accelerometer biases and gyro
drifts grow rapidly with time. The GPS/INS combination takes
advantages of each other. Velocity and direction changes from
INS output are exploited to interpolate vehicle positions during
the period of GPS signal outage, while GPS measurements arc
utilized to reset and update INS system. However, accuracy and
reliability of GPS/INS based direct geo-referencing is degraded
when GPS signals are blocked for a long period. The direct geo-
referencing using GPS/INS combination performs poor in
downtown area. When overlapping the data outputs of VLMS
Interne
that ar
existin
or dig
found
vehicle
combi
f
output
algorit
param:
grounc
horizo
are as
the bu
laser 1
DSM.
coordi
oriente
Secon
ground
coordi
adjust:
GINZ
Vehicl
which
grounc
15.9 k
data. |
rectifi
traffic
are ini
only. |
buildii
autom
the se
referei
rectify
results
existir
homo;
transf
Scanning
P.ane ^-.
Laser -
Head
Figure
config
config
Figur
(a) ge