"mined by
From the
control
f lens
rammetric
ide are
sd using
surement
recision
the die
urer in
rdinates
ey with
relative
accuracy
tion of
| mainly
now, is
1otogram-
d. The.
imple to
operate,
c rumental
|ts was
jon many
apparatus
> import
relopping
DIGITAL ORTHOPHOTO-SYSTEM FOR ARCHITECTURE REPRESENTATION
Rainer Pallaske, Werner Marten, Landolf Mauelshagen
Deutsches Bergbau-Museum (DBM) der DMT-Gesselschaft für Lehre und Bildung mbH,
am Berbaumuseum 28, D-4630 Bochum 1, F. R. Germany
ABSTRACT:
For the documentation of facade structures a special solution has been developed for the production of
orthophotos by different nonparametric rectification terms (projective transformation, bivariable polynoms
n. order, multi-quadratic interpolation). A facade is divided in different planes and sub-planes, which can
be projected independently and with freely selected transformation functions. Depending on the image
quality of the surfaces different resampling functions can be chosen. The treatment can be done
interactively within an image pyramid on higher levels by visual control. Control points and plane limits are
predetermined by an analytical plotter.
KEY WORDS: Architectural, Digital Systems, Orthophoto, System Design, Workstation.
1. INTRODUCTION
1.1. Documentation of Buildings
Terrestrial photogrammetrictechniques have animportant
rank for the documentation of buildings. The aim is an
inventory of important objects of building history and to
produce plans for planning and restauration works. The
demand is the object restitution at a sufficient degree of
details under cost-effectiveness. The document
representation utilizes line drawings or controlled photo
mosaics. The line drawing emphasize the object geometry
interpreting the architecture historical and constructive
features up to a desirable degree of abstraction. The
actual inventory of a fagade can be shown in more detail
and without interpretation by photo mosaics, as well as
the documentation of decay. In certain cases a
combination of both, line drawing and photo mosaics, is
preferable as it gives a better perception of a building.
1.2. Techniques of Documentation
Inthe past controlled photo mosaics have been produced
by rectification. The transformation was appliedto images
or to parts of images of almost plane façades. This
method was followed by the differential rectification, the
_ orthophoto technique. But even this procedure has got
its limitations, as surface structures with continual shape
are presupposed (Seeger, 1979). Strong discontinuous
shapes on façades are difficult to manage. Suitable
transformations are required, which fit to the object and
the task. Façades should be subdivided into delimited
planes and other surfaces. The respective parts can be
handled according their attributes by suitable
transformations.
1.3. Rule of Digital Rectification
This demandfor various transformations can be realized
onscanned images by digital processing. The production
of digital orthophotos in the field of architecture
photogrammetry has been shown by Bähr 1980 and
Schweinfurth 1985. Today low-cost workstations and
scanners allow economic solutions of digital
transformation (Bähr and Wiesel, 1991). Digital methods
make it possible to combine any given transformation, to
integrate differentimage sources andto use the extraction
of edges (Báhr, 1980). The following procedure shows a
digital system for practical tasks of building documentation
using partialimage transformation up to digital rectification.
A universal and pragmatic approach should be in a
position to handle any control information. The future
directs to the field of stereo-image matching.
2. CONCEPTION OF RECTIFICATION
2.1. Choi f Transformation Functions
The choice of suitable transformation functions shall be
variable for partial planes and depend on the plane form
and structure, direction of camera axis, distribution and
position of control points, accuracy demands.
The digital image processing consists of a plane
transformation and a resampling of the pixel structure. It
is favourable to choose the inverse transformation (indirect
method) and to compute the resampling from the original
image matrix. Different pixelinterpolations are at disposal
(nearest neighbour, bilinear and bicubic spline
interpolation). The choice has to result individually from
theimage contents ofthe respective partial plane, because