28 
G. CONCLUSIONS : ° 
  
; 6. 
Formal reports including conclusions were not solicited from working 
group participants. In fact insufficient time was available (participants 
received data late January 1972 with a deadline for June 5, 1972, for 
submission of results) for more than submission of tabulated results. The 
conclusions which follow represent those drawn by the author: 
l. Using uniform block size and control arrays the simultaneous adjust- 7. 
ment of photographs produced RMSE's in planimetry and heights 26 to 
68 and 20 to 42 percent smaller respectively, than were achieved by 
computing with sequential procedures (refer to Table 10). Note 
that simultaneous procedures do require more central processor time " 
(1.8 to 9.5 seconds per photo) than do sequential metbods (0.7 to Ren ic 
7.8 seconds per photo). However, .the time differential between the pii 
two methods appears to be narrowing due to improved routines for Subae 
solving large systems of equations. montc 
the t 
2. Using the near-minimum control array A (Figure 1) as a base unit, it obser 
is possible to state in approximate terms that: (a) when random mum g 
perturbations only are present, doubling the control results in 
reduction of planimetric and vertical discrepancies of 40 and 55 
percent, respectively; and (b) quadrupling the control yields 
decreases in discrepancies of „50% and -65Z in planimetry and ele- 
vation, respectively. Similarly when random + systematic perturba- 
tions -are present doubling control results in decreases in planimetric 
and vertical discrepancies of 55 percent, while quadrupling control 
yields a decrease of about 70 percent. ‘ 
  
  
3. Systematic perturbations applied to image plate coordinates produced 
significant systematic errors as indicated by statistical tests of 
the standard error of unit weight from the simultaneous adjustments, 
with one exception. Participant 9 (Tables 4, 7, 8) selected weights 
for the Y plate coordinates based on the random and systematic 
errors in the sample, and chose the weights for X plate coordinates 
so as to produce a standard deviation of unit weight for the image 
coordinate residuals close to unity. Information of this type is 
usually not available and selection of weights would be considerably 
more approximate. Hence, it is felt that the simulated systematic 
perturbations are significantly large. 
4. In sequential procedures where polynomials are used for block adjust- 
ment, 3rd degree equations are necessary to correct for systematic 
errors (Table 5, Tables 2A, 2B, 2C). Division of strips into sections 
(3 sections/strip) for block adjustment produced a substantial decrease 
in discrepancies. 
5. Use of 25 points per photograph resulted in less than a 1C% decrease 
in the RMSE in discrepancies. (Participant l1 Tables 1A', 2A', 5 and 
9.) Use of 25 points per photograph and 607% sidelap resulted in 
decreases in RMSE's of discrepancies of about 30% and 60 percent 
respectively, in planimetry and elevation (Tables 9, 5, 2A, Partici- 
pant 7).