e A coarse-to-fine strategy is often used. Typically, this is
implemented by means of image and feature pyramids.
e By means of a very large number of redundant mea-
surements, high accuracies can be obtained and auto-
matic blunder detection made possible.
The proposal presented in this paper tries to combine these
principles into an automated solution to the exterior/absolute
orientation — including both the detection and the pointing
problems — by means of large control objects (compared to
the image scale used in this investigation) derived from exist-
ing digital technical/topographical maps, which are available
by nationwide coverage. Both the detection and the measure-
ment of these control objects is done by an area based match-
ing algorithm, taking advantage of a coarse-to-fine strategy
and high redundancy. The proposed approach has, however,
restricting constraints that may be quite specific to our small
but well-mapped Denmark.
The potential benefit is to avoid signalization, surveying,
aerotriangulation and any extra equipment like GPS/INS.
This proposal will be perfectly suited for use in GIS, for map
revision, for monoplotting and orthophoto production, and
for production of thematic maps.
2 ASPECTS OF DIGITAL
TOPOGRAPHICAL/TECHNICAL MAPPING IN
DENMARK
In the mid-eighties, the Danish natural gas companies re-
quired that maps delivered to them should be on digital form.
This launched a development that has now resulted in a na-
tionwide coverage of digital maps, rural areas by TO maps
and urban areas by T2 or T3 maps.
TO maps with a oxy better than 0.7 m and oz better than
1 m for well-defined points are produced from 1:30,000 to
1:20,000 imagery. The number of object types are about
19. Of specific interest here is the fact that the road centres
constitute a coherent network.
T2 and T3 mapping were standardized from 1988. The num-
ber of object types are about 41 and about 51, respectively.
According to the specifications, oxy is better than 0.07 m
and oz better than 0.15 m, if accurate height control points
have been used. An example T3 map is given in figure 1.
In 1993, new so-called TK specifications including the former
TO-T3 maps were agreed on. The purpose was to ensure
better topological quality in the data and to ensure Z co-
ordinates of full accuracy. In the T and the TK specifications
as well there can be only one Z per XY and only one code
per object according to a priority list.
TOPI10DK is a new, fully topological map with accuracy
oxyz better than 1 m, and about 30 object types which
will be available from the national cadastral authorities in
nationwide coverage by 1999.
3 MEASUREMENT OF TOPOGRAPHIC GROUND
CONTROL OBJECTS — METHODOLOGY
3.1 Prerequisites
From the flight plan, approximate values for the orientation
parameters of the photographs are known. Xo and Yo of the
projection centre are assumed to be known within 1 cm in the
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
Figure 1: T3 map. Example from the model area.
photoscale, the flying height within 50 m, and the orientation
angles within 5 gon.
In the digital topographical/technical map for the area in
question, fully accurate Z co-ordinates are assumed to be
available.
Of course, not all object types from topographical/technical
maps are suited for area based matching. For easy target
modelling, objects on the ground should be used. For easy
and simple detection of the control objects, large structures
like roads, footpaths and parking areas should be used. If
detection and measurement of large structures is successful,
it will be possible to proceed to small structures like manhole
covers and gratings, for urban areas available from T2/T3
maps.
For this study, the object types road, footpath and parking
areas are extracted from a topographical/technical map. It is
assumed that most, but not necessarily all, road intersections
can be automatically detected in the data, and that it is pos-
sible to generate a closed polygon for each target template.
This has not been investigated yet, and some manual edit-
ing has done the job. When TOP10DK becomes available,
however, this is probably no longer a problem.
Finally, it is assumed that the parameters of the interior ori-
entation are available already.
3.2 Target templates
The target polygons are projected into the images by using
the approximate orientation parameters from the flight plan,
and a vector-to-raster conversion is performed for each image
in the following way:
1. One of the polygon points is automatically chosen to
be origo of the target template co-ordinate system. In
this way, the X, Y and Z is known for the template-
origo.
2. Pixels outside the polygon are given the value 100, and
pixels inside the value 200.
Figure
left of
Spline
points
system
solutic
points
3.3 |
A coat
mids.
to as
size is
values
the co
2195
With t
Seems
level 6
3.4 |
Them
autom
pointit
and a
and vi
progra
rithm:
