5A-3-4 
The stereo vision system consists of two high speed, 
high resolution progressive scan CCD cameras. It 
captures stereo image pairs while the system drives 
at posted road speeds. The images are taken 
according to an operator-defined distance interval to 
provide full coverage of the roadway and its 
surroundings. 
In order to perform precise positioning, it is 
necessary to calibrate the entire system. Camera 
geometry, lens distortions, relative orientation 
parameters and offset between the stereo vision 
system and the GPS/INS system must be 
determined. The camera geometry may be constant 
for a long time, but the relative and rotation offsets 
may change between different missions. Therefore, 
the entire system calibration is divided into two 
parts. The first part is the calculation of camera 
parameters using known control points in a test- 
field. The second is the calculation of the relative 
orientation and rotation offset based on a 
coplanarity equation and known constraints. In the 
second part, the position and rotation parameters of 
an image pair are used. No additional controls are 
required from the operator. It is well suited for an 
operationally smooth mobile mapping system. 
3.4 Feature Extraction From Stereo Imagery 
The GPSVision system is designed to capture 
images of road environment and the feature 
extraction software is developed to collect features 
from images. This software is based on Microsoft 
Windows 95/98 operating system and is external 
rule based driven and language neutral. The user 
will interact with the software and point at features 
of interest in the stereo image pairs. The software 
then triangulates the relative position of the selected 
feature and transfer it into the global coordinate 
system. 
In addition to calculating locations, the software will 
provide for the selection, control and display of 
images. A map window display the tracking of the 
GPSVision, the image location and the extracted 
features. An image pair is selected from the map 
window or using forward/backward image function. 
The image can be viewed in the zoom mode and its 
contrast is easily adjustable. The user selects the 
active feature and assigns attributes to it. After 
measuring the left and right image coordinate, the 
global coordinate or the distance measurement is 
also displayed. 
By using the image matching method, the features 
are positioned by selecting a point on one of the 
stereo images. It reduce the feature extraction time 
by half. Another application of the image matching 
technology is to measure the profile and surface of 
object. 
Fig 4 Stereo images are used to locate and measure infrastructure 
The output data is used for construction and inventory.