1-2-3 
knowledge base grow after each 
mission the percentage of 
unusable data will get smaller 
and smaller. This means that 
tasks currently done by a human 
expert will be taken over by the 
computer up to the point that it 
will be more practical and 
productive to re-survey a 
diminutive portion of a the 
survey than to spent time to 
achieve the perfection. 
wrra acr wvara ir | 
S» 
IO (000086 
utmn icoam» 
in pìFsFwrr fwrrtv»»! 
is pRFwïïSF (w7rci rw 
ùkv&om 
<2^ 
<5P 
'afoon ... I 
f ' ' 'or 1 Cnal | 
#38 af Picttct 
wrana.«r | 
■ 
u 
. 
s 
h 
Figure 2: VISAT Planning Module 
Figure 1: The VIS AT Expert Knowledge System 
The output of the Mission Survey module are synchronized 
GPS, INS, and digital images. The images are enhanced 
using the image-processing module then they are 
compressed and downloaded with an identifier to a server 
waiting to be georeferenced. At the same time, the raw 
GPS/INS navigation data are downloaded after passing 
through a pre-processor module. The pre-processor analyze 
the data and establish a processing strategy based on 
experience gained on previous surveys which are stored in 
the expert system. Typical examples are, omitting satellites, 
which frequently disappear, splitting long ZUPTS and 
defining those parts of the survey that may need backward 
smoothing processing. The GPS/INS data are then ready 
to be processed using the processing strategies defined by 
the pre-processor. Typically, the majority of them will pass 
the quality test after standard processing and will be 
merged with the imaging data for georeferencing. They will 
then be stored in the image Database. Those parts of the 
total traverse that do not pass the quality test are 
immediately submitted to a more elaborate second stage of 
processing. In this fully automated procedure, standard 
problems, such as those caused by lock of loss, are 
addressed and automatically resolved. After this stage, 
most of the data, say 98%, should be available for 
georeferencing. Those data which still do not satisfy the 
project accuracy requirements will either not enter the 
georeferencing stream or will be subjected to the scrutiny 
of a human expert who decides on the basis of the 
processing already done, whether or not further processing 
is likely to result in a higher percentage of usable data, then 
report that to the expert system. Since the expert 
After georeferencing and storage in the image library, the 
images can be used to generate the output requested by the 
user. This output will obviously be different from one user 
to the next. In many cases, the user will want to do the 
feature extraction himself. In that case, the georeferenced 
images are simply transferred together with a standard 
report on their quality. In other cases, the user may request 
specific products that can be handled by dedicated 
application software. In some cases new software 
development will be needed. To handle the enormous 
amount of data and to cover a wide range of diverse 
applications a structured Database Management System 
(DBMS) is absolutely essential. It must be capable of 
image selection based on location, time of survey, survey 
unit, best geometry, etc. On the other hand, utility 
programs for large groups of applications will also be 
needed. For many applications a partial automation of the 
measuring process will be highly desirable, such as the 
automatic measurement of conjugate points using epipolar 
lines or the automatic identification and measurement of 
geometrically well-defined objects, For map revision, 
features such as superimposition and back projection are 
extremely important. 
3. THE EXPERT KNOWLEDGE SYSTEM 
The expert knowledge system continuously interacts with 
the calibration, planning, survey-mission, real-time quality 
control, and post-mission quality control modules. In the 
following the expert knowledge in these modules will be 
discussed in more details.