Having said this, we need to recognise the laser scanner’s limi 
tations in data collection. The SOISIC LG scanner utilised is 
unable to register points within a space of less than 2m. In or 
der to collect 3D information from tight spaces it is necessary to 
use a different system, such as a short-range laser scanner 11 or 
by taking stereopairs. 
3.2 Metric error propagation for large vs. small objects 
In post-processing the data the following observations could be 
made. An important problem that was incurred, and which was 
partly due to largeness of the structure/site, was that the metric 
error (0.6mm) was multiplied when consolidating the large 
number of viewpoints. As a consequence, the viewpoints taken 
at this stage do not consolidate properly into one structure 12 . It 
must be noted, however, that the error can be avoided if a pri 
mary viewpoint can be taken which contains most of the target 
spheres that will then form part of individual viewpoints. This 
can easily be done with relatively small objects. For example, 
in the case of the Tarxien Temple altar (HI67cm x W 128cm x 
D 180cm) at the Malta Museum of Archaeology (Stage 3 - see 
Figure 5), the metric error propagation was negligible when 
Fig. 6 Outlines in laser scanned data are not clearly 
11 MCR was considering options from MINOLTA and OPTRONICS 
for such purposes 
12 It will be noted that at this early experimental stage no theodolite 
total station was used but instead the 3D1D team relied totally on 
the sphere targets provided by MENSI 
tape and known measurements were compared with the scanned 
data. In the case of Ggantija temple, a primary viewpoint con 
taining a sufficient number of spheres could be attained within 
one or at best between two opposite apses, however, this would 
not be possible for the temple as a whole. 
3.3 Lack of clarity in outline forms obtained through laser 
scans 
Although the laser scanner gives very rich surface detail, it does 
not provide sufficient data to construct the clearest outline pos 
sible of the object scanned. This becomes more evident where 
lines in reality are clearly defined, such as in the following ex 
ample from a laser scan of St John’s Cathedral’s crypt, Malta 
(Stage 4). 
It is immediately noticeable that the collection of surface data 
from the laser is relatively easy. Obtaining such a number of 
points at a grid of 2mm is no mean feat. All this data was col 
lected in a period of two nights. Considering the amount of data 
collected, there were relatively very few human hours involved. 
This statement is made in relative terms when comparing the 
even when smoothening (right) technique is applied 
results of the laser scan to the post-processing work required by 
photogrammetry, in which it is very tedious to collect all data on 
the surface in order to produce the mesh/3d model. Photogrammet- 
rical restitution is also very time consuming and its accuracy is very 
much dependent on the capability of the operator. 
- 99 -