5A-3-3 
For example, a stereo pair of images can be 
captured at an interval of 16 meters 
3. GPSVision Data Processing 
The Positioning procedure of the GPSVision 
consist of two steps, determining the position and 
rotation of the image pair in a global coordinate 
system and positioning an object from stereo image 
pair. The first step is to combine the GPS and INS 
data using the kalman filter method. The second 
step determines the three-dimensional coordinate of 
an object by a photogrammetric triangulation and 
transfers it into the global coordinate system. In the 
following, the positioning procedure of the 
GPSVision is presented. 
3.1 GPS Positioning 
Depending on the GPS receiver, the positioning 
accuracy varies. In the first generation of the 
GPSVision system, the code-phase sub-meter 
receiver is used. The CA pseudoranges are used for 
differential positioning. In the second generation of 
the GPSVision system, the dual frequency GPS 
receiver is used to obtain up to 10cm level 
positioning accuracy. 
Due to the inherent integeger ambiguities, the carier 
phase measurements lack the geometric strength 
required for high accuracy positioning. It is 
necessary to determine the excact interger 
ambiguities. One method is to used the wide lane 
technology [Hofmann-Wellenhof 1993, Dedes 
1995]. With the known integer ambiguities, the 
carier phase data are then conversed to wide lane 
psedoranges data and are used to calculate the 
high accurate position. 
The dynamic positioning with the wide lane 
pseudoranges is performed with two passes through 
the data. In the first pass, the wide lane ambiguities 
are estimated between each cycle slips. In the 
second pass, ambiguities are fixed and then used to 
perform high accuracy positioning. 
3.2 GPS/INS integration 
The integration of GPS/INS can be performed at 
different levels and using different methods. 
GPSVision technology benefits from the Kalman 
filter method [Gelb, 1974, Wei, 1990, Lapucha, 
1990] which consists of a prediction and an update. 
Fig. 2 shows the procedure of this method The 
state vector includes attitude, position, velocity, 
accelerometer biases and gyrodrifts. The 
measurement of an inertial system come from two 
sensor triads, an accelerometers block and a gyro 
block. They are defined as three components of the 
specific force vector f and three component of the 
body rotation rate.. 
m 
Fig.2 The GPS/INS integration procedure 
After establishing the dynamic model of the system, 
the Kalman prediction estimates the state vector and 
its covariance matrix of the system. Whenever a 
measurement is available, the Kalman update will 
use it to calculate more accurate state vector and 
covariance. This will repeat until all data is 
processed. In the GPS/INS integration, the data 
from the INS is very accurate for a short period, so 
instead of using the Kalman prediction, the INS 
positioning equation is used as the prediction 
module. To achieve the most smooth result, the 
Kalman filter is used in forward and backward. Fig. 
3 shows a data set after the GPS/INS integration. 
Fig. 3 The GPSVision creates the street 
map of Bayside in Wisconsin, USA 
3.3 Positioning with stereo images 
After the GPS/INS integration, every image pair 
taken by the stereo cameras is georeferenced with 
three position parameters and three rotation 
parameters. A three-dimensional coordinate of an 
object is calculated by a photogrammetric 
intersection procedure using its left and right image 
coordinate and then transferred into a global 
coordinate system..