\SER
INS based
butes to a
fication of
rs of each
' points of
NZA area,
it the laser
n interface
road lights
VLMS are
single-row
ind six line
unted on a
ng streets.
xtured 3D
borne laser
a), and an
ts, such as
sign/signal,
tic manner
ency of the
etails was
; in central
images are
orientation
, the geo-
s might be
1ave to be
geographic
> measures
| be heavily
urban area.
leasures the
with high
es and gyro
ation takes
anges from
ions during
rements are
curacy and
is degraded
direct geo-
is poor in
s of VLMS
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
that are taken in central Tokyo, such as GINZA area, with other
existing geographic data, such as DSM (Digital Surface Model)
or digital map, disagreement between different data sets is
found (see Figure 8 for an example), and it varies along the
vehicle's measurement course as the accuracy of GPS/INS
combination changes with the local surroundings.
This paper contributes to a method of fusing the data
output of VLMS with existing geographic database. An
algorithm is developed to correct the position and orientation
parameters at each GPS/INS update using e.g. a DSM as the
ground truth. Rectification is conducted in two subsequent steps,
horizontal and vertical registration. First, a number of tie-points
are assigned manually establishing correspondences between
the building frames that are measured by the geo-referenced
laser range data (called laser points) of VLMS and those in
DSM. Three parameters at each GPS/INS update, i.e. (x,y)
coordinates of vehicle position and yaw angle (x) of vehicle
orientation, are corrected to fit VLMS data to the ground truth.
Secondly, a vertical registration is conducted to match the
ground elevations along vehicle's measurement course, where z
coordinate of vehicle position at each GPS/INS update is
adjusted. The algorithm is tested using the VLMS data of
GINZA area, one of the major commercial centers in Tokyo.
Vehicle's measurement course lasted for about 15.7km. A DSM,
which was generated from an air-borne laser data that has a
ground resolution of 1m” and a ground coverage of about
15.9 km?, is used to rectify the GPS/INS parameters of VLMS
data. In data fusion, a set of objects are extracted from the
rectified VLMS data, consisting of commercial signboard,
traffic sign/signal, road light, road boundary and so on. They
are integrated with a 1:2500 3D map that has building frame
only. In addition, line images of VLMS are projected onto the
building facades of the 3D map to generate textures
automatically. In the followings, we will first briefly introduce
the sensor system of VLMS as well as the method for geo-
referencing laser range and line images. The algorithm for
rectifying GPS/INS parameters is presented next. Experimental
results as well as the applications of data fusing in updating an
existing map are addressed subsequently. In this paper,
homogenous notations are exploited to address all the
transformation matrixes.
Scanning Surfac
9 i f ac ivit
Mans Building Surface B,.- pisels
] ^ : x . Fish-aye lens
| "Tree Sensor Pcsilicn SF. inmFa, 120%
laser eu A Blind Area Ha B:
- E \ ¥ 3
Head arr gr re iis B ; BE
Ground Surface 8 * \
Ny = Line camere
(9 o» i
Figure 1. Laser range finder (LD-A) and line camera (a)
configuration of LD-A, (b) range points in a scan line, (c)
configuration of line camera.
©:INS/GPS
N NOU Line camera
RI #2 RD
E a M
£& uà A
V (a)
Figure 2. Conceptual figures of Geo-referencing data sources
(a) geo-referencing of range scan line, (b) geo-referencing of
495
line image.
GPS Antenna GPS Antenna
1, 7 /
Ih Ka x 3
LD-À = (C Von A T The
a df 3 Pr s
9r Jr A x NM
a: Ld
xl A ! F
TT “sk y. Là v.
— x : L^
Or Ag T Or Erg
»
World coordinate system — /' Line camera Wold coordinate system
(9 e
Figure 3. Conceptual figures of Geo-referencing data sources
(a) geo-referencing of range scan line, (b) geo-referencing of
line image.
2. SENSOR SYSTEM AND DATA OUTPUTS OF VLMS
2.1 Sensor system
VLMS consists of three different kinds of sensors and each for
a specific purpose. They are laser range scanners - the sensor
for measuring object geometry, line cameras - the sensor for
capturing object texture, and GeoMaster - the moving platform
with a GPS/INS based navigation unit.
Single-row laser range scanners, LD-A, produced by
IBEO Lasertechnik, are exploited in the sensor system (see
Figure 1(a)). In one scanning (a range scan line), LD-A profiles
480 range points of the surroundings on the scanning plane
within 300 degrees. A blind area of 60 degree exists due to the
hardware configuration (see Figure 1(b)). LD-A has a maximum
range distance of 100 meter and an average error of 3cm.
Frequency of LD-A is 20Hz, implying that it profiles 20 range
scan lines per second.
Line CCD cameras are implemented in the sensor system.
Each has a 8mm F4 fish-eye lens with a vision field of 180
degree on it (see Figure 1(c)). In each snapshot, a single-row
image (line image) of 1*2048 pixels is captured on the scanning
plane. Among the 2048 pixels, about 224 pixels ( i20" ) on
each side are discarded due to high lens distortion. Line images
are captured at a rate of 80Hz by each line camera.
The measurement vehicle (Figure 2(b)) - GeoMaster is
equipped with a high accurate GPS/INS based navigation
system - HISS (Konno et al. 2000). Three LD-As and six line
cameras are mounted on the roof of GeoMaster as shown in
Figure 2(a). Both LD-As and line cameras are installed with
their scanning planes at different angles to reduce occlusion
from e.g. trees. In this research, all exterior calibration
parameters (relative angles and distances) between sensor's
local coordinate systems are obtained through physical
measurement; all interior calibration parameters (e.g. focus
length) are obtained from maker or sensor's specifications. For
data measurement, all sensors keep recording data sources as
the vehicle moves ahead. When GeoMaster moves at a speed of
20km/h, line images are captured at an interval of about 6.9cm
by each line camera, range scan lines are profiled at an interval
of about 27.8cm by each LD-A, locations and directions of the
vehicle are measured by GPS/INS at an interval of about 20cm.
GPS/INS parameters are linearly interpolated and associated to
each range scan line and line image.
2.2 Geo-referencing data sources
Figure 3 shows the conceptual figures of geo-referencing range
scan lines and line images to a world coordinate system.
According to the GPS/INS parameters that are associated to
each range scan line and line image, a transformation matrix