iat the 
in the 
such a 
ecause 
cerned 
nternal 
sumed 
cluded 
| point 
| radial 
e were 
set of 
sultant 
lues in 
trained 
of the 
uld be 
artially 
in the 
nation. 
ilinear 
of the 
m the 
1dle 
  
Jace 
dard 
  
1trol. 
object 
target 
t were 
sidered 
odolite 
ed, the 
81 and 
p right 
|, UMK 
sets. In 
(ith the 
| using 
camera 
it film 
linates 
ing the 
bration 
proach 
igh the 
lied to 
   
(ii) Affine; in which an affine transformation (Equation 1) 
was applied globally to all image coordinate 
measurements. The parameters from the transformation 
were derived from a least squares fit of all 100 reseau 
image measurements to their calibrated positions. 
Ax-aotaix-ta2y () 
Ay = bo + bi x + ba y 
(iii) Mean Bilinear; on the basis that film type and camera 
back could be identified by visual analysis of the reseau 
deformation patterns derived from most of the frames 
produced, it was decided to adopt a method of correction 
which reflected these similarities (Robson 90). Instead of 
corrections based on each individual reseau image, a mean 
reseau deformation pattern was computed for the complete 
set of images. Bilinear equations (2) could then be applied 
based on the computed mean deformations at the four 
reseau crosses surrounding the image position of interest. 
This method potentially offered a camera, film and camera 
back specific correction which included some of the 
localised advantages of reseau imagery but without the 
need to measure every grid. 
Ax= a0 +aıx +a2y + a3xy Q) 
Ay 7 bo * bix * bzy + b3xy 
(iv) Local Bilinear; the conventional bilinear correction 
approach applied as for the mean bilinear method, but 
using reseau deformations derived from measurements of 
each individual image. 
3. Results and Discussion. 
3.1. Adjustments featuring the UMK 10/1318 camera 
Two adjustments based on the first survey datum were 
computed from the UMK photo-coordinate data. The first 
6-photo adjustment exhibited larger image residuals than 
the 2-photo adjustment (Figure 2). Conversely the twin 
photo network showed larger object coordinate 
discrepancies, but exhibited a smaller variance factor and 
photo-coordinate residuals. 
Table 2. Some parameters from the UMK 
Adjustment, Set (1) 
  
RMS Object Space 
Degrees Coordinate RMS 
Photos] of [Variance]  Discrepancies and ^ |Photo-coordinate 
Freedom] Factor | Standard Deviations | residuals (tm) 
(mm) 
X Y Z X y 
0.12 | 0.11 |-0.10 
(0.69) | (0.74) | (0.65) 
0.46 | 0.33 | 0.30 
(0.55) | (0.64) | (0.54) 
  
  
6 504 | 1.280 2.02 | 1.53 
  
2 51 | 0.630 1.33 | 0.20 
  
  
  
  
  
  
  
  
  
  
In the absence of gross errors, it must be concluded that 
significant departures from the assumed image plane were 
present in the glass plates at exposure. Such displacements 
could be due to the design of the UMK plate holders or 
due to humidity effects (Forno and Kearney 1987). 
    
Interestingly no significant differences were found 
between the estimated values of the 8 parameter camera 
functional model included in each adjustment. Therefore 
changes in the constrained camera parameters cannot be 
used as a measure of functional model applicability in a 
weak network. 
UMK 6-Photo Ad Justment 
404 405 
| / : : T s > > 
do \ d 1 A - Lnd x 
Pa D es - 1 T - - - 
m“ ~ ^ ay - = = . 
- - x ™ 1 xh 4 _— ‘ 
= ’ 153% 
t - ‘ / N E 1 
-— 4 - - es 1 47 PA 
tà A 
— 4 = 7 e ~ 1 
I0 um 10 um 
UMK @ Photo Ad Justment 
404 405 
æ - - 
T > - 9 - 
- + - 
Figure 2. Image Residuals from both UMK adjustments. 
Each image measurement has an associated standard 
deviation such that in the twin photo adjustment the image 
space is better defined than the object space. For this 
reason the majority of residuals from the least squares 
process have been pushed into the estimated object 
coordinates. Conversely the multiple photo network has 
provided strong constraints in the object space by virtue 
of its geometry, as a consequence the residuals tend to 
remain in the image space. 
UMK 6-Photo UMK 2- Photo 
Front View / 
di + 4 = - NOS ~ | 
1 x E 2 ec M 
: i 
= r ^ — x / — Y 
im ab bii ci Ai à 
X x + a 3 EIU = Co en 
: = ; z d i / E ici I 
- 5 i : / <= ut 
= xt 1 1 FE = 
pH 4 
2 mm 
AIF TIT. Y E) 
" # = 
Figure 3. Object Space discrepancies from both UMK 
adjustments 
The systematic object space discrepancies (figure 3) show 
that significant departures from the imaging geometry 
assumed by the functional model have occurred in the 
UMK camera. Such problems are based on holding the