CIP A 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey
2.2 Laser Scanning
In recent years, laser scanning has found increasing application
in 3D documentation ( Barber 2001, Boehler 2001 Fangi 2001,
Lingua and Rinaudo, 2001). When the opportunity arose to use
a laser scanner, it was decided to explore the potential of this
technology for documentation in Africa, where access to sites is
generally difficult and site condition can be harsh and adverse
to the use of sophisticated technology. The Kilwa ruins, for
example, are located on an island without access to electricity,
and batteries had to be transported by Dhow to the mainland for
charging. Further, the delicate and heavy equipment had to be
transported over badly pot-holed roads and earned by hand in
difficult terrain.
Figure 3 Scan of part of the pillar hall. The pillar scan
comprises of 230 000 scan points
In six days of scanning, 180 scans were recorded totalling
approximately 15 Million scan points or 2.1 Gbytes of data. The
number of daily scans was limited by the battery life of the
system computer and the need to recharge the battery overnight
Figure 4 . Laser scanning facing the badly damaged
inside walls of the Gereza
on the mainland.
In a comparison of laser and close-range photogrammetry
techniques, based merely on the experiences of the Kilwa field
campaign and a few laboratory tests, one can summarise first
impressions as follows (see also Boehler et al, 2001).
The principal advantages of laser scanning over close-range
photogrammetry emerging from the work on the Kilwa
documentation are:
The Cyrax 2500 laser scanner, available for the 2003 July field
campaign, has a recommended range of 50m but scans up to
100m are possible under good conditions. The system’s
horizontal and vertical scan-field-of-view is 40° and a
maximum of 1000 scan points in each direction can be recorded
in one scan. The laser dot size is 6 mm over 50 m. Quoted
standard deviations are a = ± 6 mm at a range of 1.5m - 50m.
Laboratory testing prior to the fieldwork confirmed this
accuracy, with maximum total point error of 11 mm over 20m.
Structural detail, edges and targets can be scanned in sub-scans
with higher resolution and added into the coordinate system of
the master scan. The fortress and mosque were recorded with
scan point intervals ranging from 10 mm to 25 mm at distances
of 5 m to 15 m from the features scanned, depending on detail
of the recorded structural component and spatial constraints.
Figure 4. Scan of a dome shaped ceiling inside the
Great Mosque of Kilwa. The 2m by 2m by 1.4 m
dome was recorded with 130 000 scan points at a 10
mm scan interval.
laser scanning provides direct and immediate access to the
scan data making it possible to visually inspect the point
cloud in situ and identify possible problem areas in the data
sets in the field.
the point cloud is obtained without any additional
processing. Post-processing is similar to that for
photogrammetry.
at an accuracy level of 10 mm, no obvious outliers such as
surface spikes, typical for point clouds derived from image
matching, were observed in the scan data. This is contrary
to experiences with close range scanners reported elsewhere
(Lingua and Rinaudo, 2001).
only one set-up is required for each surface. This'saves
significantly on planning and execution time and is
especially advantageous for complex interior rooms. (In the
case of the Kilwa documentation, a hall with some 20
narrowly spaced columns was recorded
photogrammetrically in an earlier field campaign. Finding
suitable camera positions to capture multiple images for all
column surfaces proved extremely difficult and the data
acquisition was time consuming. Capturing the same
surfaces with the scanner (Figures 3 and 4) required only a
fraction of the time and effort compared to the
photogrammetric data acquisition.
Close-range photogrammetry on the other hand appeared
superior to laser scanning in the following aspects:
close range photogrammetry provides discreet user-selected
points. Vector data, edges, comers and decorative detail can
be easier identified and extracted from images than from a
point cloud. When using a laser scanner, this can be partly
overcome by high-resolution sub-scans of relevant detail.
