The size of an interpretation overlay is limited (23 
x 23 cm) and the content is not very complex. Often 
it is only a single theme and a single line type. A 
typical landuse interpretation consists e.g. only of 
polygons that are not overlapping and which have a 
single attribute. Each piece of land can only belong 
to one polygon (one landuse). Automatic digitizing 
of such an overlay should be relatively easy. 
4 COMPARABLE APPROACHES 
MEISNER (1981) from the Remote Sensing Laboratory of 
the University of Minnesota has been using a medium 
cost image processing system to obtain 640 x 480 
pixel video images of map sections. An operator 
positions the screen cursor inside a polygon and 
enters the (landuse) code for that polygon. The 
computer then scans from the indicated point tó the 
left until a boundary is reached. An automatic line 
following routine is used to follow this boundary 
back to the starting point. The entire process takes 
1-2 seconds for a typical polygon. 
LEBERL (1982) describes the use of a rather 
expensive flatbed scanner, the Karto Scan, for the 
automatic digitizing of polygon data. He used the 
scanner only for scanning the polygon boundaries 
only. Centroids with polygon labels have been 
digitized separately on a manual digitizer. The 
scanned boundary data are vectorized, using Syscan 
software, (before scanning, labels had to be removed 
from the map in a predigitial preparation step). All 
further processing is in vector format. 
The geometry of the information as digitized is the 
same as the geometry of the airphoto that has been 
interpreted. In certain cases it is possible to 
digitize overlays from orthophotos where the 
geometry is already correct, but orthophotos should 
not be prerequisite for the system. 
The overlay will be "push button" scanned in a 
raster mode, results are pre-processed and displayed 
on a high resolution color CRT. Next step is to 
indicate with a lightpen or cursor on the CRT lines 
or areas with identical codes and to enter those 
codes. That should be easy since all codes on the 
manuscript are also scanned and hence displayed on 
the screen. 
An alternative approach is to digitize the centroids 
of each polygon on a conventional digitizer together 
with its code and use this datafile for automatic 
coding. 
When a code is assigned to an area the software will 
look for its boundaries and paints the area with the 
color allocated to that code. The colored map on the 
screen enables easy checking by the interpreter. 
Editing, (deleting, changing or entering areas and 
boundaries) can be done either by correcting the 
original overlay drawing and redigitizing it, or by 
interactive editing at the terminal. 
When the colored map is correct, it may be sent to 
the appropriate database after applying the 
specified geometric correction(s). 
5.2 Hardware 
photodiode 
up to 12 b 
computer, ; 
state arra; 
properties 
photolithoj 
1981). Eiki 
pixel, cori 
verified, 1 
its digita 
ITC includi 
compensate 
a computer 
separation 
(Another E 
The CCD hoi 
sensitive 
manufacture 
his earliei 
Using the ! 
overlay mei 
per mm or i 
mm line wii 
effective •, 
5.3 The cui 
Fig 3 gives 
currently < 
used for A1 
11/24 and < 
bitplanes. 
At the 1986 ACSM conference N.R. CHRISMAN reported 
on the use of a scanning system, incorporating a 
Data Copy 1700 x 3800 pixel CCD camera. The system 
is used to digitize 9 x 15 inch soilmaps. After 
thresholding and thinning vector output is generated 
that is further processed with a "spaghetti and 
meatballs" approach and edited to obtain a 
topologically correct datafile. Scanning and 
vectorizing a soil map took 20-30 minutes. Chrisman 
believes that "devices like this scanner will alter 
the digitizing environment to create higher quality 
with lower expense". 
Also recently published has been a description by 
FAIN (1985) of a commercial system, the Smart Scan 
System, developed by Energy Images Inc., that seems 
to use (according to the accompanying illustration) 
an Eikonix 4096 x 4096 CCD array scanning camera. 
The Smart Scan system outputs vector data and uses 
some artificial intelligence software to divide the 
vector data into various layers based on information 
how lines and symbols are represented on the 
document. Subsequent editing is carried out in an 
interactive vector graphics environment. The scanned 
raster data are displayed on the CRT screen with the 
derived vector data superimposed, so that errors can 
be detected easily. 
5 THE ADIOS CONCEPT 
5.1 How ADIOS will look to the user 
It is assumed that the result of an interpretation 
will be in the form of a (transparent) 23 x 23 cm 
overlay, belonging to a particular airphoto, drawn 
with a rapidograph pen ofO .3 mm linewidth or 
larger. To a limited extent it may be possible to 
impose certain conventions to the drawing style of 
the interpreter (cf norms for technical drawings 
that have to be microfilmed). That means that the 
interpreter should draw the fiducial marks in a 
certain way, mark tiepoints where applicable, keep 
text and lines separate, close or overshoot polygons 
etc. 
At present three technologies should be considered 
for a simple system 
- videodigitizing 
- digitizing with photodiodes or CCD’s electronic 
camera 
- low cost document scanners 
Typical resolution for video digitizing is 512 x 512 
or 640 x 480 while for photodiodes or CCD’s 2048 x 
2048 resp. 4096 x 4096 may be obtained. 
For the exi 
used (fig. 
1980 for a 
They are tj 
were indeec 
automatic c 
results ob1 
représentai 
The various 
The low cost document scanners introduced in 1986 
have resolutions of approximately 2500 x 3500 , but 
little is known of their spatial accuracy as yet. 
They only handle up to A4 size originals and, like 
photocopying machines, they lack a viewfinder for 
orientation of the original. Some systems have the 
capability of handling color. Table I summarizes 
some characteristics. 
TABLE I: A4 Image Scanners Hannover CEBIT fair 1986 
Type 
resolution dpi 
grey 
levels 
color 
OCR 
Ricoh IS 30 
300,240,200,180 
16 
- 
_ 
" CS 30 
" 
ft 
RGB 
- 
" FS 1 
400,300,240,120 
256 
RGB 
- 
Compuscan PCS 
200,150,100 
7 
- 
+ 
Canon CS 220 
300,200,150,75 
i6 
- 
- 
dpi = dots per inch ocr = optical character 
recognition 
MEISNER, already mentioned above, used a SPATIAL 
DATA EYECOM II Picture digitizer, a video digitizer 
with 640 x 480 resolution. 
Applied to a 9 inch photo that gives resolution of 
0.4 mm which is hardly sufficient if a 0.3 mm 
rapidograph is used and is considerably lower than 
the resolution of manual digitizers (0.1 to 0.025 
mm). 
For ADIOS, an Eikonix 78/99 digital camera has been 
selected with a linear 2048 element photodiode array 
mounted in a motor driven stage (fig.2) that moves 
in 2048 steps equal to the distance between the 
photodetectors (15/im). The voltage of each 
Step 1 scai 
In scanninj 
to éliminai 
on the ovei 
interpretit 
values of 1 
completely 
edge enhane 
Step 2 edge 
Although tl 
displayed < 
pixels are 
- unequal 
- varyinj 
- varyinj 
In the pres 
photographe 
incandescei 
found that 
Some of the 
those of se 
areas on o1 
unprocessee 
(dark linei 
thresholdii 
connectivi1 
noise, (fij 
To "reconst 
enhancement 
factors