1-2-2 
sensors considerably reduce the data processing effort by 
eliminating the digitizing step. They also opens the way 
towards new and flexible designs of the processing chain, 
making ample use of mathematical software tools readily 
available. Precise navigation has developed to a point 
where it can provide the solution of the exterior orientation 
problem without the .use of GCPs or block adjustment 
procedures. Since results are available in a digital form, 
data fusion with the imaging data is easy and real-time 
applications are possible in principle. Combining these two 
developments, the concept of the georeferenced image as 
the basic photogrammetric unit emerges. This means that 
each image is stamped with its georeferencing parameters, 
namely three positions and three orientations, and can be 
combined with any other georeferenced image of the same 
scene by using geometric constraints, such as epipolar 
geometry or object-space matching. This is a qualitatively 
new step because the georeferencing parameters for each 
image are obtained in a direct way by independent 
measurement. 
In the following, common features in the design and 
analysis of an expert knowledge system for mobile 
mapping application will be discussed and illustrated by 
examples. The main objectives of the development of the 
expert knowledge system is to design an intelligent MMS 
that speeds up the process of arriving at the required results 
and to automate all processes that requires human expert 
knowledge and interaction. 
To illustrate the major steps, the development of the 
VISAT system will be taken as an example. It is installed 
in a road vehicle, typically moving with a velocity of 50-60 
Km/h. The three main sensor subsystems are GPS, INS, 
and a cluster of eight video cameras. While the first two 
provide position and attitude for the system, the third one 
images the surrounding environment at each exposure. The 
system is synchronized by PC real-time-clock which is 
corrected every second by the Pulse Per Second (PPS) of 
the GPS receiver clock. 
2. DATA FLOW OPTIMISATION AND 
AUTOMATION 
Data flow optimization and automation are on the one hand 
based on the mathematical description and the integration 
model of the system; on the other hand, they are completely 
separate from it. Before addressing optimization and 
automation, the quiet assumption is usually made that the 
underlying mathematics of the process is well understood, 
but that the process of arriving at the results is too slow and 
requires too much human interaction. The emphasis in this 
step is therefore on speeding up the process of arriving at 
the required result, including all essential parameters that 
describe its quality, and on the automation of all processes 
that require human expert knowledge and interaction. Very 
often, the automation process is the more difficult one to 
accomplish because the further it goes, the more complex it 
becomes, and the likelihood that it will show a curve of 
diminishing return is very high. It is therefore not 
surprising that complete automation is rarely achieved, but 
that a reasonable level of automation is defined which will 
cover most of the cases that occur with a certain frequency 
(Schwarz and El-Sheimy, 1996). 
The data flow of the VISAT MMS expert knowledge 
system is shown in Figure 1. At the top level of the data 
flow are the Project Editor, Map Generator, and the 
Calibration Modules. The Project Editor includes the 
project parameters and the user requirements. The project 
parameters include project area and time allowed for the 
project while the user requirements include type of survey, 
accuracy, reliability, image coverage, result presentation 
(maps, reports, digital output), etc. These parameters are 
used to allocate the suitable resources for the project and 
then passed to the Planning module along with the history 
of previous surveys which is stored in the expert system. 
They are then used to optimize the survey. 
The Map Generator Module is used in georeferencing 
raster/vector maps and digital images that cover the project 
area. The output of the map generator is a tiled database of 
maps and/or images which can be accessed simultaneously 
according to the resolution required by the user. The 
calibration Module includes the determination of the 
cameras inner and relative orientation parameters The inner 
orientation parameters define the internal geometry of the 
camera. The relative orientation parameters define the 
relative location and orientation between the camera cluster 
and the navigation sensors (GPS and INS). The relative 
orientation parameters will be used in the transformation of 
the 2-D image coordinates into the 3-D world coordinates 
in the georeferencing process. 
The input to the planning module are the project properties 
and the user requirements defined at the top level of the 
data flow. These information are critical for the survey 
planning. They determine the selection of the survey route 
according to parameters such as satellite availability, sun 
direction and elevation, road type, tree coverage, buildings, 
speed limits, traffic density, spacing between exposures, 
length of survey, time schedule, equipment used, etc. To 
facilitate the set-up of an easy-access survey database, the 
survey route is divided into small units which essentially 
follow the road pattern or other easily identifiable features. 
The result of optimizing all these factors is expressed in a 
mission file that defines the survey trajectory and the 
operational constraints. The survey route can be planned 
using any sort of georeferenced media, produced from the 
Map Generator Module, such as digital vector maps, 
digitized aerial photo, and satellite imagery, see Figure 2.