Satellite imagery is not the only form of 
digital imagery available. Airborne digital 
scanners are available and can produce 
imagery with higher resolution than that 
available from satellites. Such imagery 
could be used to map for cultural features 
in urban areas which require more detailed 
interpretation. This would still allow the 
source data to be in a more useful digital, 
multispectral form. 
PROBLEMS 
To date digital satellite imagery has only 
seen limited use in a mapping context. 
Problems such as the availability of high 
resolution imagery, accurate correction 
techniques, and a general lack of 
understanding of digital image processing 
have hindered the entry of digital satellite 
imagery into the map production environment. 
Until the recent launch of SPOT, the highest 
resolution digital satellite imagery 
generally available has been the LANDSAT-TM 
imagery, which has a pixel resolution of 30 
metres square. Evaluation of TM imagery 
(Welch, Jordan, and Ehlers, 1985) has shown 
that imagery can be geometrically corrected 
to an RMSxy error of less than 20 metres. 
The SPOT satellite is the first commercial 
satellite capable of producing high 
resolution imagery which can be used to 
derive planimetry and topography at scales 
of 1:50,000 and smaller (Welch, 1985). SPOT 
will use two sensors, a panchromatic sensor 
(PLA) with 10 metre pixel resolution, and 
three-band multispectral sensor (MLA) with 
20 metre pixel resolution. 
SPOT will also provide the unique capability 
of producing stereo image pairs through the 
use of pointable imaging sensors. These 
stereo image pairs will make it possible to 
extract elevation information directly from 
the satellite imagery. Prototype systems 
have already been developed to automatically 
extract digital elevation models (DEM) from 
stereo image pairs. Using simulated SPOT 
data, RMS errors of less than 10 metres have 
been achieved for automatically generated 
digital terrain models (Cooper, Friedmann, 
and Wood, 1985 ) . 
The use of precision corrected, geocoded 
imagery in the map making process will 
dramatically improve the quality and variety 
of 'cartographic products that can be 
produced. Simply put, geocoded imagery is 
imagery which has been transformed to a 
desired map projection with the rows and 
columns of pixels aligned with the 
projection axis. In fact, geocoded imagery 
provides more than just transformed rows and 
columns. Geocoded imagery has also been 
corrected for all source dependent errors, 
resampled to a standard square pixel size 
(e.g. MSS pixels are resampled to 50 
metres, TM to 25 metres, SPOT MLA to 12.5 
metres, and PLA to 6.25 metres), and each 
geocoded image is sized to an integral 
number of standard mapsheets. 
There are numerous benefits to be had from 
using geocoded imagery. The two most 
obvious include satellite sensor 
independence of the data and easy 
integration with other image and non-image 
data. Because geocoding removes source 
dependent errors and transforms the imagery 
to a standard map format, the problems 
associated with multi-source images are 
eliminated. Imagery from both LANDSAT TM 
and MSS sensors and SPOT can be easily 
combined and manipulated together. 
Similarly, geocoded imagery can be used in 
combination with any other data that shares 
the same map projection. 
General acceptance of mapping from satellite 
imagery will only come about once its 
feasibility can be proven. This paper will 
attempt to show that mapping can be done 
from such imagery through the use of new 
digital techniques and that it is feasible 
to map from satellite imagery in an 
operational environment. 
THE MAP PRODUCTION PROCESS 
The production of the sample topographic 
base map from digitally-processed satellite 
imagery followed the flow shown in Figure 1. 
FIGURE 1 - The Map Production Process 
The key step is the precision image 
correction and geocoding procedure. This 
procedure allowed the raw satellite imagery 
to be corrected to the cartographic accuracy 
standards required by the map maker. The 
combination of geocoding and correction in 
this process made it possible to place the 
digital imagery into a common map coordinate 
system without additional accuracy-degrading 
transformation steps. 
This technique involved modeling the motion 
of the spacecraft and sensor during the data 
collection orbit. Through the use of a 
sophisticated spacecraft model, ground 
control points were used only to "fine tune" 
the model for more precise geometric image 
correction. This allowed the use of many 
fewer ground control points to achieve 
subpixel accuracy than would be required for 
standard warping methods. In general, only 
7 to 15 control points are required to 
precision geocode a 34,000 sq. kilometre 
image to subpixel accuracy. 
The extraction of elevation information 
required a stereo pair of corrected images 
as input. The elevation extraction process 
is fully automated. The computer matched 
corresponding features in the left and right 
images and determined the relative elevation