ere 
ken 
ving 
ling 
her 
ing. 
ing 
rms 
ma, 
by 
ude 
tch. 
y & 
pro- 
Iser 
ori- 
, for 
hich 
ight 
was 
t al, 
lied 
ose 
king 
hilst 
ereo 
jach 
ised 
rule- 
AEX 
pro- 
ly in 
per- 
data 
nent 
age 
Sys- 
ated 
ape 
and 
2. MIS DESCRIPTION 
MIS is similar to NAMEX in that it is a hybrid Prolog-based 
system where high level queries and problems can be ex- 
pressed in a logic program of English-like Prolog predi- 
cates, facts, and primitives. This hybrid nature allows the 
benefits of Prolog’s ability to manipulate high level sym- 
bolic data and its inference mechanism (backtracking) to 
be realised whilst utilising the data processing efficiencies 
and fast numerical analysis capabilities of low level proce- 
dural languages, for which Prolog has no particular current 
advantage. Low level functions are accessed via Prolog 
primitives calling low level procedural routines written in 
FORTRAN77 and Kernigham and Ritchie ‘C’. Presently 
(May 1992) MIS consists of approximately 1000 lines of 
Prolog, 50 lines of C, and 1000 lines of FORTRAN code. 
The low level routines which perform data processing and 
numerical analysis are largely implemented in FORTRAN. 
The C component of the low level procedures is used as 
a communication interface between the Prolog primitives 
and FORTRAN, and for specifically selected tasks involv- 
ing the high speed processing of raster data. The kind of 
tasks performed by the low level routines include camera 
modelling; detection of control point containment within 
images; the computation of image centres and corners; a 
model of the spheroids of Mars (Wu, 1981) and Phobos 
(Duxbury, 1991); overlap detection between two or more 
images; and the scanning of specified regions in a raster 
based-map. 
Most of the data utilised by MIS was obtained from NASA 
CD-ROMs (NASA, 1988, 1992A), and include a gazetteer, 
a l' per pixel geological raster map of Mars, and Viking 
Orbiter camera event and geometry data. Additional data 
was supplied by the USGS Flagstaff in the form of a 
file containing 4505 USGS control point locations (Wu & 
Schafer, 1984); a set of improved camera event and ge- 
ometry data (Wu et al, 1982) and a + per grid-point DEM 
for the whole of Mars (NASA, 1992C). Pre-processing of 
all of these data was necessary to convert them into a form 
suitable for MIS. The gazetteer contained feature names, 
locations and sizes — these were converted into Prolog 
facts. The DEM data had to be resampled to a standard 
1° per pixel raster map format. A new Viking Orbiter im- 
age geometry and event file was also generated by com- 
bining the Viking Orbiter CD-ROM data with the smaller 
set of USGS improved data. The USGS records replaced 
corresponding CD-ROM records and were given a higher 
quality value. 
2.1 Prolog Primitives 
A Prolog primitive is a basic command which performs a 
single specific operation via low level procedural routines. 
Prolog primitives control these low level routines by pass- 
ing command and control codes, and data, via a central 
command routine. The command routine then instructs 
the other routines to read or write data, or to perform data 
processing, or calculations. The outcome of a command is 
returned to a Prolog primitive either as a flag to indicate a 
successful completion or as data which will subsequently 
be used by Prolog. The use of Prolog primitives effec- 
tively shields the user from the complexities of the low 
789 
level procedural routines. 
An example of a Prolog primitive is: limb(/mage). This 
takes an image name, passes this to a low level routine 
which then identifies the appropriate image record and 
calculates the rays through each corner of the image. If 
any corner ray fails to intersect with the planet then the 
primitive succeeds and the limb of the planet is visible. 
Another example is: stereo_illumination(Image1, Image2, 
Value). This passes the names of two images, which over- 
lap, to a low level routine which then computes the solar 
altitude and azimuth for each image. The altitude and 
azimuth information is then used to calculate the ends 
of shadow tips from a unit high object on the surface of 
Mars at the centre of each image. The value returned 
(x10), known as the “stereo illumination difference”, is the 
unit difference between the tips of both shadows. This 
value was used as an approximation as to how suitable 
stereo image pairs are, illumination-wise. A large illumi- 
nation difference value indicates that shadows present in 
both images will have changed significantly and as a result 
the stereo matcher will fail. 
Most of the Prolog primitives which call low level routines 
are generally concerned with images and features. Some 
of these primitives make use of other Prolog statements 
to perform simple high to low level data translation tasks, 
and vice versa, during interaction with low level procedural 
routines. Appendix 1 lists a small selection of the Prolog 
primitives used by MIS. 
2.2 Prolog facts 
There are two classes of Prolog facts used in MIS, "com- 
prehensive" and "example" facts. "Comprehensive" facts 
are where a complete set of facts are available such as 
all the named features on Mars. "Example" facts apply 
to cases where the information being presented is incom- 
plete and the facts available are only representative. This 
latter type of Prolog fact is prefixed by the word "example". 
Approximately one thousand cartographic features are 
stored as Prolog facts and each contains a feature name, 
location, size, and feature class. Circular structures such 
as craters or volcanoes are represented by centre loca- 
tions and a diameter (km). Linear and area features are 
currently represented by longitude and latitude limits of an 
enclosing bounding rectangle, and a length (km) e.g. 
feature('Copernicus', 169.0, -50.0, 280, crater). 
feature('Olympus Mons', 133.0, 18.0, 540, volcano). 
feature('Ma'adim  Vallis', 184.0,181.0,-28.0,-16.0, 955, 
sinuous valley). 
It was decided to utilise the existing USGS map sheet 
and quadrant system notation (Greeley & Batson, 1991) 
to specify default regions in which stereo matching would 
take place. The fact map_limits contains the map sheet 
number, the quadrant, and the map's longitude and lat- 
itude limits. If required, a new user defined non-USGS 
map sheet area can be defined by asserting an additional 
map.limits fact e.g. 
map_limits('Olympus Mons area’, map, 128,140,12,25). 
Te PS Eee