CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey 
Algorithms for automated or semi-automated feature 
extraction from 3D point clouds, combined with appropriate 
point cloud thinning methods overcome or reduce this 
shortcoming of the laser scanner. 
photogrammetric point position accuracy is typically highqr 
for targeted natural points at short distances (5m to 15 m). 
At larger distances the accuracies of photogrammetry and 
laser scanning become similar and at distances approaching 
100 m laser scanning appears to provide slightly higher 
accuracies. 
cameras are significantly lighter, easier to transport and 
mechanically more robust than laser scanners, 
photogrammetrically acquired photography provides 
permanent records, allowing originally unplanned 
measurements of detail at a later stage 
for highly textured surfaces, point clouds can be generated 
at higher densities than laser scans. In laser scanning this 
disadvantage can be overcome by repeat scans with slightly 
modified orientations. 
at present, photogrammetric equipment is significantly less 
expensive than laser scanning equipment, 
photogrammetric procedures are designed to provide 
redundancy (Barber et al., 2001), while the laser might 
tempt the operator to accept data from a single scan. For the 
Kilwa documentation redundancy was achieved for the 
principal surfaces by repeat scanning of the same surface 
from a different station. 
No difficulties were experienced with poor surface reflection. 
This can be attributed to the homogeneous surface structure of 
the Kilwa buildings, which consist of well reflecting light coral 
stones, partly covered with a highly reflecting lime plaster. 
At present, the integration of photogrammetry and laser 
scanning can be seen as a reliable and accurate method of 
documentation. One can, however, anticipate that the above 
listed advantages of close range photogrammetry will be 
equalled by scanners in the near future. One can expect laser 
scanners to be equipped with fully integrated, stable, high 
resolution digital cameras, in a configuration which can be 
calibrated to create a common geometry for the integrated 
system. One can, indeed, venture to predict that laser scanning 
will have the same impact on close-range photogrammetry as 
GPS had on conventional surveying. 
3. DATA PROCESSING 
The Kilwa documentation has the dual objective of recording 
the structures in Kilwa as well as the development of an 
appropriate methodology suitable for documentation of African 
heritage sites in general, bearing in mind restrictions in 
technical and funding resources and limitations due to difficult 
environments. Three approaches were chosen for processing the 
acquired imagery and laser scans. 1 
1. Sequential processing -photogrammetry employing a 
combination of in-house and off-the-shelf software. 
2. Integrated processing -photogrammetry employing a stand 
alone integrated software programme. 
3. Hybrid processing- combination of laser scanning and 
photogrammetry. 
3.1 Sequential Processing 
The photogrammetric processing follows standard procedures 
of photogrammetric triangulation (using Australis software), 
line drawing (AutoCAD), image matching (in-house software,) 
ortho-image generation (in-house software) and 3D modelling, 
as for example described in Riither et a, 2001). 
3.1.1 Data structure for automated photogrammetry-CAD 
transfer. A new model for the data acquisition with Australis 
software was developed with a view to an automated transfer of 
photogrammetrically digitised feature points into a CAD 
system. The following workflow was adopted: 
An approximate CAD model was created using tape 
measurements of the Gereza’s ground plan (Figure 5). 
Figure 5. A bird’s eye view of Gereza’s preliminary 
solid model designed for planning purposes 
The working model was used to label (code) each point 
according to its location within the structure, the feature 
type and the location on the feature, according to a pre 
designed data structure; for instance, EG1-1 (Figure 6) 
refers the first point on the left side of the gate in the East 
wall, or EWPW1_1 refers to vertex 1 of window 1 on 
Figure 6. East wall and detail of entrance gate 
showing an example of the labelling system for 
automated point identification in CAD