I —— 0——^ 
| Li) Li) | L6) I IL(g-D| i —QNW—— 
  
  
  
  
   
   
  
     
  
   
  
   
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
     
  
  
  
        
    
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Cu) C(2) C(2) em C(3) C(4) C(4) C(5) Cg: 1) Clg) I 
i 4 i 1 1 4 4 ! : 
fü) (S) | L] 
Swi J Pe pi Pe der) x =] 
9% 9 94 35 9 «—r(3) (s) | | 
& {| 
= 12) L ni 
By nhà ih dn d----- p o e 7] | 
| -—r(4) = 2) 
-— r(3) (S) C5) i 
S(3) IN r(4) à ) 
wig] d» 8 9 O5 D 4 I 9 
— (5) i | i (Le [3 
| ; -i | | % S— —  D-—— mM 
| | I : IL | L(2) | L(3) | L(4) 1 LS) | pe —W— ei 
| à Bub i 
-— F(S) (g-!) | - 
sol dh-dl dh 4-4 dese e. [P2 | | | | 
-— ríst2) 9- i E I i 
i i He i Pa a 4 dl ue 
Figure 4.2. Arrangements of models and tie points. hl. i =I rT 
  
S-2) 
  
LLLA 
pA pe (g-1) 
La) 
wR | FRE 
Sn (g-1) (—=G 
1 
EA l 
L(2) | 
—-]|9 g+! | |g+2|--—-—— 2(g- — 
Lee 2 
| 
| 
+, 
  
  
    
BC 
  
  
  
     
B(S) 
   
d 
  
  
L(s-3 
| 
| | 
| 
2 + = i 
E Figure 4.6. Matrix patterns, p=q=60% . 
  
  
  
  
  
  
| 
l 
| 
| 
  
  
  
  
  
  
  
  
Ls Se (+131 2)-———— en I "n a The ordering of models in this case was tried using 
— i ; E un different strategies, namely: (2G), (^S), diagonal 
( — 6) | = = $i front, Cuthill-Mckee method for minimum bandwidth 
: E - zi (Cuthill, 1972) and spiral front. The ordering 
Figure 4.3. Ordering of models. (—S) according to these strategies is demonstrated in 
figure 5 for a network of g=5 by s=3. The minimum 
number of F.I. as well as the minimum bandwidth had 
been achieved by ordering the models either (2G) or 
(+S) pending the dimension of the network (see 
conditions for economic ordering in table 1). The 
b(k,k) patterns according to these orderings are presented 
c in figure 6. It should be noted that the resulting 
patterns are the same and the differences are only 
in the dimensions of b and the number of B 
bk) constituting M. 
  
IE ] |! Like) 
I6-01 12 e 
El L| 
D(k,k+ 2) 
es If the fore-lap is increased to 80$ the models from 
LA consequetive photographs in a strip would be con- 
structed with a base = 20%. This base would give 
[7 intersections in the model space of less reliability 
if compared with the intersections produced from a 
[E b base - 40$ (figure 7-1). Therefore it might be 
E] ; advantageous in this case to construct the models 
ii 
= 
  
  
  
  
  
  
  
in a strip from every other photograph (figure 7-2). 
Figure 4.4. Correlation windows. Moreover this approach resultsin half the number of 
successively constructed models from a strip, which 
leads to economy in both observations and com- 
putations. It shall be assumed, therefore, for all 
the cases of p=80% that models from a strip are 
Pg) | (g-) 1g) | constructed from every other photograph. Figure 8 
  
  
  
  
(S) ., 
li represents the patterns for ordering (?G)and (^S). 
=80%, q=60% 
(S) N (g-1) " N 4.5. p=80%, g=60% 
| blk,k) | blk,k+1)! D(kk+2) The patterns corresponding to this case are 
=~ | | illustrated in figure 9. 
b(k,k) b(k,k 1) (—- 6G) 
( ——9 S) 4.6. p=80%, q=60% (4C) 
o 
  
  
  
  
  
  
  
  
  
  
Figure 4.5. Matrix sub-block Bk) In this case the cross models are introduced. The 
244 
S() 
— 
S(2) 
e 
S(3) 
m 
(i) Dov 
L(1) 
Le 
[3] 
(ii) Acr 
[7] 
[3] 
(ii) Dia 
[a] 
[4] 
(iv) Min 
/ 
m 
* 
(v) £ 
Figur