1€ 
;h 
  
taken with the mount inverted and held down through the 
smoke holes from the roof (Fig.4). All photographs were 
taken on Ilford FP4 film which is a consistent and tolerant 
emulsion. Processing was again carried out ın a hostel 
bathroom using Ilford ID11 at 1:1. 
No UMK photography was obtained in 1990, but about 
160 photographs were taken with the Hasselblad. As in 
the previous year, about 450 survey measurements were 
made. 
4. THE BUNDLE ADJUSTMENT. 
An account of computations which were carried out 
following the 1989 and 1990 visits is given by Biddle et 
al (1992). 
The primary survey data consists of photographs, 
archaeological descriptions, sketches and plans and the 
site survey measurements. From these primary data, 
secondary data can be derived by different techniques, in 
different forms and for different purposes. One of the 
major purposes of the survey is the accurate spatial 
definition of the main structural features of the Edicule. If 
it had been possible to take stereopairs of metric 
photographs, each with more than three well disposed and 
well defined coordinated control points, then accurate 
photogrammetry could have been carried out directly. 
Given the site problems, some photographs, particularly 
of the spaces above the level of the lamps inside the 
Edicule, included no predefined control points and some 
had only one or two. Thus a bundle adjustment was 
            
   
CN 
RENS 
RS 
y 
E =4) X 
$a 
x 
e, 
— 
AV; 
1/7 
N DS 
sx. 7 
ZA 
N 
X 
LA 
AYA) 
N 
  
= S 
    
   
  
   
  
  
  
   
  
  
Hi v 
"AN 
[WT 
E SUA 
‘= 
iS 
AN 
     
     
OS 
UN 
Y y; 
Figure 5. Cut-away linestring model of the Edicule (in progress). 
  
necessary to provide a homogeneous set of coordinated 
points for each stereopair and exterior orientation 
parameters of all the photographs. 
    
    
   
  
  
     
   
  
  
  
     
   
   
    
    
   
  
  
    
     
    
      
   
   
   
  
  
   
Pass points (or tie points) were identified and marked on 
prints. The photocoordinates of these and of the surveyed 
targets were measured in an Intergraph Intermap Analytic 
(IMA). These measurements were used in conjunction 
with the site survey measurements of height differences, 
horizontal angles, zenith angles and slope distances in a 
linearised least squares process to estimate the spatial 
coordinates of the survey stations, the surveyed control 
points and the pass points and the exterior orientations of 
the previously calibrated cameras. 
The computation was carried out using the Generalised 
Adjustment Program (GAP) written in C by staff and 
research students of the Engineering Surveying Research 
Centre. The program was run under UNIX on a Sun 
Sparcstation 1+. 
Memory storage limitations (16Mb + 50 Mb swap space) 
meant that a simultaneous solution for all 2418 unknowns 
was unacceptably slow, taking about a day for one 
iteration; almost all the run time was used for swapping 
data. The dense network of photography and survey did 
not lend itself to automatic rigorous sequential estimation, 
so a quasi-sequential method was used. Two sequences of 
independent observations were used, but only the 
variances of the coordinates estimated in the first 
sequence were input into the second sequence; 
covariances were ignored. Later, a full simultaneous 
H— E 
N 
< 
t] 
ll 
i 
© 
5] 
      
   
   
  
  
   
  
   
AZ 
zu 
N 
pon 
7 
7 
A 
3 
      
  
zz} 
j 
| 
NS 
/ 
S 
SN 
= 
D» 
| 4 
Ras ] X 
y SNO 
de) 
JN 
SN NS 
^ d 
Lj 
      
  
E—— 
HO LN 
A 
SN 
i 
Y 
UN 
A 
\ 
V/ 
/ 
[A 
X 
/ 
  
IV. 7 
(FAN) 
p 
[| 
  
Ü 
  
ic. 
V 
7 x) x 
VA 
b