ty 
to 
of 
383 
4. ANALYSIS AND VISUALISATION OF DEFORMATION. 
Each panel was removed from the easel, stored and subsequently repositioned for measurement twenty four times. As 
a consequence, the position and orientation elements of each datum were not consistent. Two methods were applied to 
solve this problem. 
4.1 Rapid visualisation. 
The first method, applicable for convenient and rapid visualisation purposes only, was based on a rigid body 
transformation. The four targets situated in the corners of every panel were located and used as common points to 
position and orientate each panel with either its location at epoch 0 or at any other epoch. A NURB surface was fitted 
through these data using Intergraph Microstation. A direct visual comparison could then be easily obtained by 
overlaying and rendering views from any pair of epochs. It was particularly useful to render the surface from one epoch 
in a semi-transparent manner such that changes above and below could be directly seen from a single isometric 
viewpoint (Fig.6). 
Rapid Measurement: Panel 308. 
Top Horizontal Centre horizontal 
RM Epoch 2r ATA T 
a ee 
me Epoch 10 7 me 
= mm any 
fia Epoch » T9 
MM Epoch 4 
Fig. 6 Visual surface difference 308 epochs 7 and 17. Fig. 7. Edges for panel 308 rapid test (10x exaggeration). 
  
An advantage of NURB surfaces is that profiles and sections can be easily extracted to provide information which is 
readily understood by the end user. Fig. 7. demonstrates the extraction of a series of profile comparisons for panel 308. 
Profiles, taken from the horizontal top edge and a line parallel to it but level with the panel centre, have been extracted 
from the top left quadrant of the panel. It is easy to see from this sequence how the panel shape has altered during a 30 
minute period in which humidity was changed from 40% to 80%RH. Code is currently being written to produce 
automatically a similar, but three dimensional display. 
4.2 Use of the same object co-ordinate starting values followed by global and local congruency testing. 
The simple visualisation techniques described are by no 
x ; \ means rigorous since they contain no information about 
en tes. data quality or an analysis of which points or regions have 
satt | t eee: \ actualy moved or are stable. A rigorous method of 
t. die. determining deformation is to evaluate the co-ordinate 
| differences between epochs in conjunction with their 
. ; ; | associated covariance matrices. Initially, a global 
tet oo . eo d e congruency test (using a=0.05) is carried out to determine 
A " / whether any significant movements have occurred 
between epochs. If the global test indicates significant 
x e movement, it is necessary to locate the unstable points. At 
the same time, it is necessary to re-define the datum for 
the computation (with respect to the stable points). The 
localisation procedure used in this case is modified from 
Fraser and Gruendig = (1988), and consists of 
e eC 
Fig. 8. Glass master deformation vectors at 10x 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995