S per 
tures 
gital 
f the 
aring 
sible 
Point 
tion. 
s was 
s on 
aphs. 
e (PC 
ength 
a of 
, Was 
jects 
"^ the 
tized 
is the 
| the 
'esult 
'tting 
ition, 
1-made 
itural 
laries 
linear 
gross 
less 
1rious 
)urces 
tizing 
psilon 
ed to 
ne; it 
most 
sition 
t from 
line 
ring a 
s true 
and is 
line's 
psilon 
quires 
alysis 
ind the 
il the 
nd or 
remain 
d. 
y, and 
n with 
I. The 
rmatted 
the PC 
rom the 
verage. 
ine was 
tandard 
s also 
THE EXPERIMENTS 
Materials 
Having topographic database revision in mind, we 
selected the following materials for the 
experiments. 
Wide-angle photographs (diapositives and paper 
prints) from the Gould area (southern France) at 
scale 1:30,000, taken in 1989 
Wide-angle photographs (paper prints) of the 
same area at scale 1:30,000, taken in 1976 
IGN (France), topographic map at scale 1:25,000 
of the same area, photogrammetricaly produced 
from photographs taken in 1980 and revised in 
1986 
Digital map, produced on a Zeiss C120 analytical 
stereoplotter from the photographs taken in 
1976, and stored in DGN Microstation format 
Analytical plotter 
There are several analytical plotters on the 
market today. Some have very sophisticated designs 
and high performance, but are very expensive in 
terms of both investment and maintenance. There 
are also low-cost analytical plotters with 
somewhat simple designs and using the popular 
budget-priced PCs. Their precision is lower than 
the sophisticated ones, but may be the best choice 
for some specific applications. A survey of 
low-cost analytical plotters can be found in [6]. 
The Topcon PA-2000 analytical plotter was used in 
the experiments. It was designed at ITC (The 
Netherlands) and is licensed to Topcon. The 
instrument has one photo carrier for the two 
photographs. The photo carrier is movable in X and 
Y directions and rotatable around a fixed axis. 
One rotary and two linear encoders connected to 
the photo carrier are use to determine positional 
and angular coordinates relative to a fixed 
coordinate system. 
The PA-2000 incorporates a unique concept for the 
inner orientation of the photographs. The film or 
paper print has to be perforated by a punch tool 
that matches the corresponding studs on the 
instrument’s photo carrier. The orientation 
procedure consists of only relative and absolute 
orientation. The inner orientation is obtained by 
preparing the photographs as described above, and 
measuring the central points. 
The magnification of the optics is 4x (6.5x with 
an optional eyepiece). The resolution of the 
measurement system is 5 um, while its measuring 
accuracy is 20 um. The instrument at ITC supports 
Microstation PC as a 3D or 2D digitizing software. 
The system software is designed to run on PCs with 
the following minimum recommended configuration 
- IBM-AT compatible 80286 processor and 80287 
coprocessor 
- colour monitor VGA, 16 colours 
- RS-232 communication port 
- tablet with 4-button cursor (Calcomp drawing 
board 2300 series) 
- GPIB-interface PCII/IIA or RS232 
Data collection 
The data needed for the experiments were divided 
in two groups. The first group contained the 
495 
reference data, and the second group contained the 
test data sets. 
The reference data for attribute accuracy 
evaluation were derived from the IGN topographic 
map (scale 1:25,000). An area shown in figure 1 of 
8.8 x 11.0 cm at the map scale was selected and 
features of interest were digitized using ILWIS 
software developed at ITC [10]. 
  
  
  
INN 
LA E 
  
1 
rnnt 
| {Han 
  
  
  
  
  
  
  
Figure 1: A portion of IGN topographic map, used 
as a true attribute value. 
The available digital map was used for positional 
accuracy evaluation. 
The test data were collected from the 1989 
photographs by both digital monoplotting and 
stereoplotting systems after the necessary 
orientation data were created. The available 
control point field (seven points) was densified 
by aerial triangulation. A small block consisting 
of three models was measured on the Zeiss C120 
analytical plotter, and adjusted by the  PATM 
program. Thus 30 extra control points were 
established. 
The digital monoplotting system also requires a 
digital terrain model (DTM) for the transformation 
of feature data, sampled on the image plane, into 
the terrain coordinate system. DTM data from two 
models covering the test area were collected on 
the Kern DSR1 analytical plotter, on which 
progressive sampling software (COPS [13]) is 
installed. DTM data with 50 m grid spacing were 
generated by the SCOP program and then transferred 
to ILVIS. 
Because point features such as towers, monuments, 
windmills, etc, could not be identified from the 
photographs, and isolated houses and small 
villages had been subjected to cartographic 
generalization on the reference data set, point 
features were omitted from the attribute accuracy 
evaluation. Only line and polygon features were 
used.