5B-3-4 
DGPS positions and INS positions and the 
differences between DGPS velocities and INS 
velocities. Then-the estimated errors are fed 
back to INS solutions. If the GPS signals are 
not available, the navigation mode is switched 
to INS/Odometer mode. The autonomous 
odometer is substituted for DGPS, thus the 
odometer velocity is used as the measurement 
corresponding to DGPS. Since the DGPS and 
odometer data are collected at 1 Hz rate, the 
updating interval is one second. This 
configuration will make the inertial system to 
provide accurate orientation and navigation 
continuously. The velocity of vehicle is 
calculated from the angular velocity of rotation 
and the radius of wheel. The radius of wheel 
infinitesimally changes due to variations in tire 
pressure, frictional wear, and weight of the 
body. Moreover, the error of slips appears easily. 
So, these errors can be regarded as the 
variation in the wheel radius and defined by 
the scale factor error (5 s. The velocity measured 
by the odometer shows the vehicle’s velocity in 
the perpendicular direction of the axle-tree. 
However, the axle-tree direction does not 
parallel to the body frame. Therefore it is 
assumed that the direction is slightly 
misaligned from the body frame by small 
attitude angles (5 0 about azimuth axis and 
6 0 about pitch axis. On this basis, a linear 
differential equation describing the 
propagation of INS/Odometer errors was 
developed. The state vector to be estimated is 
as 
follows. 
1 
6 X 
x position error 
2 
y position error 
3 
<5vx 
x velocity error 
4 
6vy 
y velocity error 
5 
(5 x 
x tilt error 
6 
<5y 
y tilt error 
7 
6 z 
z tilt error 
8 
6 h 
altitude error 
9 
6v z 
z velocity error 
10 
6 li) 
odometer azimuth error 
11 
6 S 
odometer scale factor error 
12 
6 6 
odometer pitch error 
13 
bx 
x-axis accelerometer bias 
14 
by 
y-axis accelerometer bias 
15 
bz 
z-axis accelerometer bias 
16 
dx 
x-axis gyro drift 
17 
dy 
y-axis gyro drift 
18 
dz 
z-axis gyro drift 
3.3 Position and Image Input System 
In this system, all the image sensors are 
interfaced to an industrial PC (Pentium200), 
which has 3 PCI buses and 17 ISA buses. It 
inputs the revolution pulse signals from the 
encoder and generates a trigger at constant 
distance intervals. This trigger signal is output 
to two cameras (FOR-A:HMC-1170) and DGPS 
receiver unit simultaneously. This mechanism 
is required to synchronize the 3D position and 
attitude with the stereo image in post 
processing. The stereo camera, which has the 
resolution of 1280(H)xl024(V) pixels, is fixed 
on the top of the vehicle as looking ahead and 
takes pictures consecutively. The specially 
developed frame grabber board converts stereo 
video images to digital images in real time, and 
their stereo images are compressed and 
recorded to tape streamer (Quantum:DLT- 
7000,35GB) with time-tag information. The 
processing mentioned above can be executed 
within every 5 meters interval at speed of 100 
kilometers an hour. And also it is capable of 
collecting required data for three hours 
continuously. Besides, two video cameras to 
capture surrounding images are installed. 
These images are recorded into a digital VTR 
(SONY:DVCAM) with the time-tag. These 
cameras can be fixed at arbitrary points on the 
roof. Moreover they can be panned, and zoomed 
by remote control. 
3.4 Data Processing System 
The hardware comprises a desktop PC, players 
of tape streamer and digital VTR. And software 
integrates a Geographical Information System 
(SIS ver.4.1), the digital map 2500 (Spatial 
Data Framework), a database management 
system (ORACLE8) and a digital analytical 
photogrammetry software (ADIMS). The image 
data acquired by the measurement vehicle are 
constructed as 3D database corresponded to the 
navigation data. Namely, by designating 
arbitrary location on the road from the display, 
the corresponding road alignment information 
or images are searched and displayed. If 
necessary, the position, length, or shape of the 
facilities can be measured. Besides, this system 
has the function of real-time simulation of the 
street scene. 
4. TEST AND RESULTS 
The test run to evaluate the fundamental 
performance of HISS was conducted in Atugi