CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey
Eastern wall, parapet 1. Further points are labeled in a
sequence which follows the outline of the window in
clockwise direction, and named EWPW1_2 and so forth.
These labels, which identify individual points uniquely, are
allocated to points during the 3D digitising with the
Australis software. In previous digitising processes, points
had only allocated numbers and manually linked in CAD.
The identification of individual points in a large densely
populated point cloud proved extremely difficult and
sometimes impossible. The new coding method combined
with automated transfer into CAD, proved a significant
improvement on the previous approach and the features of
the Gereza’s East Wall were successfully plotted on
AutoCAD with this method.
The resultant file of labelled coordinates was then translated
into a corresponding Lisp feature file, by means of a
program (Aus2lisp) for the automatic generation of
AutoCAD feature descriptions from an ASCII xyz-file.
Once imported into the CAD system, the points of a feature
are automatically joined with lines according to the number
sequence of their labels.
3.1.2 CAD Modelling. CAD environments have become
standard platforms for the display of photogrammetric results.
Since CAD models conveniently represent photogrammetrically
acquired data in a structured form with an unambiguous
topology, CAD and photogrammetry naturally blend into a
powerful tool for data capture and representation. The challenge
in this area is the need to improve the efficiency
photogrammetrically acquired data can be transferred into a
CAD system for mapping and visualization. The relationship
between CAD and photogrammetry and different approaches to
their integration are reviewed in Van den Heuvel, (2000). As
described above, the CAD model for the Gereza was created by
importing the photogrammetrically generated labelled feature
points modified with Aus21isp from the Australis ASCII format
into Lisp code. The data structure labels then inform the CAD
system on the automatic generation of lines connecting feature
points into walls windows and doors.
3.2 Integrated processing
In order to evaluate different processing techniques for use in
the recording of African heritage sites, images were processed
with PhotomodelerPro40, an integrated stand-alone modelling
software programme. This approach has the advantage that
relatively little skill is required to produce a surface model, a
fact which can lead to the well-known ‘black box’
phenomenon. A non-expert operator may well ignore basic
photogrammetric principles, accept poor image acquisition
geometry and be unable to interpret accuracy information
correctly. As a result of this, a model may well be accepted on
the basis of its appearance rather than its metric accuracy. An
experienced operator, on the other hand, can generate an
acceptable line model and, with some reservations, a texture
model. The author’s experience with such integrated software
showed that good result can be achieved in cases where the
physical environment allowed the acquisition of images in
normal orientation (optical axis at right angle to surface)
supported by a significant number of oblique images. In this
configuration, the normal case images provide the texture for
the ortho images, and the oblique images robustness and
accuracy for the photogrammetric triangulation. In all cases,
where PhotoModeler was employed by the authors for texture
modelling, a small section of the surface texture had a distorted,
“smeared”, appearance in spite of extensive attempts to improve
the model by adding surface points and selecting different
images for the areas in question.
In the opinion of the authors, programmes such as
PhotoModeler have an important role to play in heritage
documentation, especially in environments with limited
resources, but can generally not be seen as equivalent to full
digital photogrammetry work stations.
3.3 Hybrid processing
First experiments with a combination of laser scans and
photogrammetry are promising and it is planned to adopt this
approach as a principal modelling method for the project. The
laser scan point clouds were exported from the scanning
software in ASCII and dxf-format and these formats were
imported into CAD software without complication. Point clouds
can be rendered in AutoCAD and models can be created.
Problems encountered with the Kilwa scan data arise primarily
from the difficulty of merging scans in areas without transfer
targets or well defined feature points, which can be used for the
3D transformations of scans into a single model. A further
complication arises from the difficulty of working with the
large data sets, generated by laser scanning, on standard
desktop computers.
4 ONGOING AND FUTURE RESEARCH
Present research within the project is directed at three aspects of
laser scanning, namely:
optimisation of the merging of separate scans into a single
model
development of algorithms to semi automatically generate
edges and features from point clouds
generation of ortho images through the combination of
• camera position and orientation from
photogrammetric triangulation
• point clouds DSMs generated by the laser scanner.
• feature vector data semi automatically extracted from
the point cloud
developments of algorithms for data thinning
Laser scanning software does include functions for the above
tasks; however, present cost of such software make it
prohibitive for use in the limited-resources environment of
Africa. The authors also believe that there is scope to improve
some of the existing 3D feature extraction algorithms and to
add new ones. It is also planned to further explore 3D
modelling approaches and to extend the project to the
neighbouring Songo Mnara. Finally a Historical Information
System will be developed and populated with relevant
historical, environmental and site management data for the
entire Kilwa complex.
4. AFRICAN HERITAGE DATABASE
The Kilwa project is part of the larger initiative of creating an
African Architectural Heritage Database (Rüther, 2002). The