The Accuracy of Features Positioned with the GPSVan 
Guangping Hel, Kurt Novak? , Wei Tang? 
lCenter for Mapping, The Ohio State University 
1216 Kinnear Rd, Columbus, OH-43212, USA 
“Department of Geodetic Science and Surveying , 
The Ohio State University 
1958 Neil Avenue, Columbus, OH-43202, USA 
Abstract 
Data capture is the most expensive part of establishing a Geographic Information System (GIS). A Mobile 
Mapping System the <GPSVan>, integrating a stereo vision system, Global Positioning System (GPS) and 
Inertial System, has been developed to quickly and accurately collect data. The vision system takes stereo 
images, while the GPS and inertial systems provide the position and rotation of the vehicle. Any object 
which appears on an image pair can be located in a global coordinate system. In this paper, we present the 
mathematical model of camera calibration, the correction of the offsets between different components, as 
well as the analysis of the positioning accuracy of the GPSVan. 
KEY WORDS: Accuracy analysis, Camera calibration, Data capture, System integration, GPS. 
1. Introduction 
The creation of a geographic information system 
(GIS) requires an enormous amount of digital 
information. To date, most land-related 
databases still rely on existing line maps which 
are manually digitized. In order to collect digital 
data faster and more accurately, a mobile 
mapping system <GPSVan> was developed at The 
Center for Mapping of the Ohio State University 
(Bossler, 1991). 
This system consists of three major 
components: 
* the control module 
* the positioning module 
* the image module 
Control Module 
A PC-compatible computer controls the data 
collection module through a real-time, multi- 
tasking operating system. During field operation, 
a color-touch screen displays the sensor status. 
The touch screen is also used to note and record 
attribute information on objects such as bridges 
and street signs. 
Positioning Module 
A Navstar Global Positioning System (GPS) 
receiver determines the global location of the 
GPSVan. For standard road mapping applications, 
480 
the GPSVan uses single-frequency GPS receivers 
in differential mode. Depending on the type of 
GPS receiver and the operational techniques 
used, position accuracy can range from three 
meters to one centimeter. 
Because obstructions such as bridges, trees, 
tunnels or high-rise buildings can interrupt 
satellite signals, a dead reckoning (DR) system 
supplements the GPS receiver by recording the 
distance traveled and the direction of the 
GPSVan. The DR system consists of a vertical 
and directional gyroscope and wheel counters and 
accurately maps the position of the GPSVan in 
the absence of satellite positions for distances 
longer than a mile. In addition, it provides the 
rotation angles of the GPSVan. 
Image Module 
The image module consists of a stereo vision 
system, Super VHS color video cameras, and a 
data collection software package. 
The stereo vision system consists of two, fully 
digital, high resolution CCD cameras (Kodak DCS) 
with a sensor of 1280x1024 pixels. It acquires 
image pairs covering the road environment in 
front or in back of the GPSVan. The stereo vision 
system is an accurate tool for positioning 
objects in three-dimensional space. By applying 
photogrammetric triangulation techniques, any 
point th 
reconstru 
The prer 
system 
paramete 
the relati 
pair, as 
stereo vis 
Image p: 
mobile m 
Through 
the posit 
coordinat 
from the 
the uniq 
following 
* Ca 
al 
KIN 
g 
Fast, acc 
major pur 
the follov 
GPSVan 
accuracy 
2. 
The calit 
camera ( 
rotation 
calibration 
which in 
different 
control px 
image coc 
to obtait 
orientatio 
using cons 
2.1 Cam: 
The calibr 
bundle a 
parameter 
calculated 
the colline 
are define 
point (xp, 
distortion 
radial dist 
and two 
distortion |