both LANDSAT TM
can be easily
:ed together.
;an be used in
iata that shares
j from satellite
ibout once its
rhis paper will
Ding can be done
le use of new
: it is feasible
nagery in an
PROCESS
Die topographic
:essed satellite
Dwn in Figure 1.
MIC PRODUCTS
OPTIONAL
ANCILLARY DATA
tion Process
recision image
procedure. This
:ellite imagery
jraphic accuracy
jp maker. The
id correction in
; to place the
i map coordinate
:uracy-degrading
Ling the motion
during the data
the use of a
model, ground
r to "fine tune"
geometric image
: use of many
its to achieve
be required for
in general, only
:e required to
) sq. kilometre
in information
:orrected images
action process
romputer matched
> left and right
.ative elevation
from the
Wood,
informati
digital
regular g
The elev
contours
software.
(Cooper, Friedmann, and
From this parallax
computer generated a
model of the surface on a
required map projection,
d was then converted to
for output using DEM processing
parai
lax
1985)
.
on,
the
eleva
tion
rid i
n th
ation
gr
Base map pi
precision
a combinati
inte rpretiv
feature ext
involved t
multi-spect
Planimetric
were highl
methods.
were output
digital fi
were then u
orthophotos
planimetry
the Survey
the Brit
Environment
animetry was derived from the
geocoded image using two methods:
on of image analysis and manual
e techniques; and automated
raction techniques. The first
aking advantage of
ral nature of the
features (water,
ighted using image
These digitally enhanced images
to film, with a high resolution
lm recorder. The film products
sed to produce 1:50,000 scale
and interpreted to derive
by trained photo-interpreters at
s and Resource Mapping Branch of
ish Columbia Ministry of
the digital
TM imagery,
roads, etc.)
enhancement
A more automated approach for extracting
feature planimetry was also tested. This
technique used an automated knowledge-based
classifier approach to identify the water
and road information based on both spectral
detail and such knowledge as size, shape,
and context. Once these features were
identified they were transferred into the
GIS and merged into the map data base.
Thematic information can be derived from
digital imagery by computer classification.
The results of this technique, however, are
rarely usable in a mapping context without
some form of post-processing. This is
required to give the resultant image a more
map-like appearance and to convert the data
into a format that can be manipulated on a
GIS (i.e. polygon). The use of
knowledge-based filters (which use actual
cartographic rules), ancillary data, and
raster-to-vector conversion processes allow
the production of actual thematic maps from
the imagery.
The planimetric, thematic, and digital
elevation information derived from the
satellite imagery was stored in a Geographic
Information System (GIS) data base. The GIS
provided a data base which can store and
retrieve information by geographic location,
and a set of facilities to mainipulate and
analyze geographic data.
The output map products include the
topographic base map, thematic maps, and
orthophotos with graphic overlays. All of
these can be derived entirely from digital
satellite imagery. Also, the digital nature
of the source material allowed for the
production of other non-standard map
products. For example, by combining the
digital elevation information and the
digital satellite imagery it was possible to
produce such products as three-dimensional
perspective and panoramic views.
EVALUATION OF THE SAMPLE MAP PRODUCT
Figure 2 shows a sample 1:50,000 scale map
which was created from LANDSAT-5 TM imagery
using the techniques described in the
previous section. This map covers a section
of northern Vancouver Island, Canada, known
as Adam River. It corresponds to the
Canadian NTS 1:50,000 scale mapsheet 92L/8
published by the Surveys and Mapping Branch,
Department of Energy, Mines and Resources.
The map was derived from overlapped stereo
LANDSAT-5 TM imagery obtained from adjacent
orbits (Path 49 and Path 50). The dates of
the source imagery were July 15, 1984 and
July 14, 1984. The overlap from adjacent
orbits range from almost none at the equator
to full overlap north of 52 degrees.
For the sample map, 7 ground control points
were used to precision geocode a scene
covering an area of approximately 34,000 sq.
kilometres to sub-pixel accuracy. This area
is equivalent to that which is covered by
sixteen 1:50,000 scale mapsheets.
Conventional photogrammetric methods would
require a minimum of 80 ground control
points to cover an equivalent area and
precision image warping techniques may
require over 50 points.
FIGURE 2 - Sample Satellite Derived Base Map
To calculate the planimetric accuracy of the
resultant image, control points which were
not used in the correction process were
measured and compared on both the resulting
map and the original map. For this map the
calculated planimetric error was determined
to be 27.3 metres RMS.
The determination of the elevation accuracy
involved comparing the LANDSAT image-derived
DEM with a reference DEM (produced by
photogrammetric means from air photos by the
British Columbia Ministry of Environment).
The comparison showed the LANDSAT
image-derived DEM had an RMS error of 72.0
metres when compared to the reference DEM.
LANDSAT imagery has a small base-to-height
ratio (about 0.1) due to the small
difference in the viewing angle between
adjacent orbits. This limits the elevation
accuracy which is attainable from LANDSAT
image ry.
SPOT can produce stereo imagery by pointing
its sensors up to 27.5 degrees off nadir to
view an area previously imaged from a
different orbit. Stereo imagery can be
produced for any area, with a base-to-height
ratio of up to 0.5 (if one image is a nadir
view) or 1.0 (if both imaqes are viewed at
the maximum off-nadir view).
The accuracy of the elevation information
derived from stereo satellite imagery
depends on the pixel resolution and
base-to-height ratio of the imagery.
Currently available LANDSAT-TM stereo