PHOTOGRAMMETRIC ENGINEERING 
from adjacent segments (J, k) and (j+1, 
k) using matrix notation to designate the 
segments. 
It should be noted for both control points and 
tie points that: 
Xin) — Bj. po T T = Cj 
5" 
/ \ 
Uu) 
- (4) > (à) 
AXjk = AA CX; m 
and 
eay?- T FB +R 
(2) 
AY 
where: X Y are the mean coordinates of 
the corner tie points and control points for 
segment (7, k). 
Since there is a redundance of relative equa- 
tions (note equation (6) ), expressing the re- 
lationship between segments (generally 12 
equations per segment) and normally more 
than two control equations (note equation 
(5) ), a least squares solution results from 
minimizing the sum of the squares of the 
residual errors; that is, 
n 
sS ( po^ + > (V, à? + > (V te) 
il il 
ium] 
TO t)G:uo,) - minimum 
The method had been tested on a project 
area with ten strips of 16 to 20 exposures each. 
The photo scale was 1:18,000, and there were 
eight control points around the perimeter of 
the area. There were 86 segments, 344 co. 
efficients for the secondary transformations, 
16 condition equations on control and 602 
condition equations of relative differences. A 
total of 36 geodetic control points was used 
for checking. The results of this test show that 
the RMS error on these points was 12.25' with 
a maximum error of 27’. It is felt that these 
errors have been adversely affected by errors 
in the input data and the method should 
normally give better results. With the con- 
ventional independent strip adjustment, em- 
ploying the procedures normally used at the 
Army Map Service, it is estimated that 22 
control points, instead of 8, would have been 
required for obtaining the same results. The 
time with the UNIVAC for this 
adjustment was sixteen hours; however, as 
Computer 
previously mentioned, existing programs were 
used. It has been estimated that with the pro- 
gram which is currently being written for this 
particular problem the time will be signifi- 
cantly reduced. 
  
  
Archi