from 
on has 
ly the 
larger 
s have 
dinate 
ffered 
urious 
-photo 
  
Best results have been obtained from the locally applied 
bilinear correction. This is not surprising since the method 
is both local and specific to each frame, providing the best 
estimate of the actual image deformations occurring. 
Unfortunately the method is also the most measurement 
intensive, each complete set of reseau cross images for 
every frame included in the bundle having to be measured. 
The radial and tangential parameters estimated by each 
adjustment were also significantly different within this set, 
implying that the parameters were also modelling some of 
the uncorrected film deformation. Again the image 
deformations have had little effect on the estimated object 
coordinates because the object space is highly constrained 
by the convergent network. 
3.3.2. Results from the twin photograph adjustments 
with the Hasselblad camera. 
All adjustments based on two photos have demonstrated 
that higher object coordinate precision requires localised 
correction provided by individual reseau measurement. 
Non local correction of image deformation may be ignored 
with this camera. This is especially true where scale 
corrections based on fiducials are concerned since large 
deformations, due to unflatness and tension at the format 
edges, can contribute disproportionately. The raw 
photo-coordinates have provided reasonable results with 
a minimum of both computational and measurement 
effort. 
Hasselblad 2-Phorto 
Full Control Partial Control 
4 Raw Plate 4 à 
* + 
QM eee 
+ + + + + * 
S dibus irat | en | 
| A + + + |. * + + . 
pére tan 
fA] 7 lu 
X 4 + 
i um 10 um 
4 Local Blllnear 
xx , ; ; = 4*1 * : + 
* ata x * : | TR 4 : + | + 
Î ot s + I A odes + 
+ + 
pr ed # 
: ^41 i 
f | #4] Me 
+ + x + + 
n. on 
Figure 6. Some image coordinate residuals from the 
stereopair Hasselblad surveys. 
0b Ject Space Discrepancies 
Front View 
Affine Local BllInear 
EN EN pe any 
d dore rd 
My TUN eue 
/ 
AER M id 
Top View 
Row Plate Affine 
AY MAT fe 
Mean Bllinear 
Local Billnear 
M? nw 
1 X puis 
Figure 7. Object space discrepancies for the twin photo 
Hasselblad survey using datum 1. 
0b Ject Space Discrepancies 
Front View 
N Aff Ine Local BilInear 
3 / / N E / à nd 7 
foi | E 
yrs / V, sb fif 
/ ^A + \ | À > = ] 4 
wl TN rl 
f BV P spa E 7 
) > x 1 Y e - 
ri mm 
J View 
Raw Plate Affine 
E Mean BllInear Local Bl/Inear 
NN A = > 
Figure 8. Object space discrepancies for the stereopair 
Hasselblad Survey using datum 2.