s.
EE TETE EEE Torre
k^
rescented in some detail in this section in
table 2.
ANALYSES OF RESULTS
The results listed in table 2 are quite
useful in finding answers to basic questions
related to experimental design. The analyses of
results, in general, follow this scheme.
Bi. Does MOV technique improve accuracy of esti-
mate of Y and how does it compare with simple
Monte Carlo technique?. To begin with, the re-
sults nf case À and case H may be compared. The
reduction in variance is of the order of sos.
Additional computer time required for this is
only about 252. The answer is thus clearly affi-
rmative.
. Which statistical scheme/eg. is best
suited to the normal case in close range photog-
rammetry? We can find the answer in this
through step by step comparisons. Result of
cases B and C indicate that it is better to use
the same random number stream for processing the
nces at the various points and further that
preferable to adjust the random numbers to
a mean of zero. Again, comparison of case
pr case À with case D indicates that it is
heiter to use a computed value using eq.3.4 far
bi vector rather than use a value nf ! for each
of the elements in the vector. If we compare the
results of cases D or F nr B against result of
£
Varia
its
visid
we + =
case H, we understand that use of eg. 3.5 a &
combined in one would be preferable to use c
Bither 3.5a alone or 3.6b alone or e&eg.3.1. This
is attributable to the maximum reduction in the
magnitude of the residusics realizable in rase H.
er
“ta
Table 2: Variantes obtained by use of different
methods
EERE RIBERA eat an ni Eté en ds
B. pi.d.pl.2.pio DESCRIPTION/RENARKS
.5 5,7 11.6 Simple Monte Carlo tech. 85=
53. No HCY Para. used. Fun
time = 2M 125.
Resind used, File:MCVOIR OUT
B 112,4 90,3 37.8 Œhi=i, HCV technique used, MCY=
B.Different ran.numberz used
at each step. Rand. nos. nnt
adi. to mean ©, All MEY can
be located in pone plane. Ho
rewind used, HCVOIE,QU7.
Ref.eg.s,1
C 103.5 79,0 62,4 MEV tech. NEW z B.Hand. nos.
adj. to mean 0, hi-i.Ptz.iü8,
19, 235,22, 31,358.37 | reveal
large diff.bet.calc. & true
values, All MEV Ÿ are equal,
No Rewind HCVOIC.QUT,
Ref, eg. 3.1.
D 2.5 5.7 11.3 MLV tech. 58254. MDCV-B,. Five
pts chow neg cov. bet Y k X.
Correction computed as in
Kobayashi.Ref.eg.3.1.Var. is
not reduced. MHUCVOID,DUT.
m
Hal
Ln
rn
^
42
bt
Ea
^
dt
n
e
^
rn
Fa
^
t4
trs
^
kJ
{Table ? Continued.!
HCY tech.B points all in one
plane. Diff, rand. nos. used
for Bach control variable.
MCVBID.2UT. Ref.eng. 3.1.
HCY tech, 55=54,WHCV = 8. In
computing corrections cont-
ributinns arising from pos.
bi values only are taken
into account. Yer. red, i=
quite good, MCVBIE.DUT.
Ref. eg. 3.&5b.
t
2.1 5,7 HCY = 8. They are unequai Y
coord. of control points,
Yar.red. obtained by use of
neg. bi values only.
HECVOIF.DUT.Ref.en. 3.81.
"1
e
x
RI
H 0.04 0,4 T.& HCV = B, bi values calc.
from CEBINV, Var red ohtd.
by consideration and seper-
ste treatment of positive
and neg. bi values.Run time:
2M 455. HDCVGiB5.üuT.
Ref. eg. 3.483 & 3.8kh.
i 358.0 28,8 21.0 MOY = 3eB =24.No improvement
in var. of comp.Y-coords.S5=
54, Fun time:4H105 HCVOZR. OUT
Ref, eg. 3.88 & S.àh.
d 1.25 3,5... 8.2. MEV= 4.0nly Y-zoord of cont,
pts used.Var.red, = 20%. The
4 pts are situated similar
tn mid-side fiducial marks
in HCVB3A.0UT.The Four pis
arg situated at corner nf
tont. field in HCVUSA.2UT.
Run timez2M 375.HOVEIA. GUT &
.BSUT.Ref. bg. 2.53 & 3.àb.
i 0.7 2.8: 7.0 MCV = 7. Bignificant red. of
var. Run time-z2M 308.HCVB3A
.DUA,.BUR,, GUC, BUD,
Ref. eg. 3.58 & 3.&b.
p
i283 4.1. 9.1. MEN e 1. Binnificant red, of
var, is observed. Run time =
2M 238. MCVOGA.QUI T8 .8US.
Ref. eg. 3.83 & b.
H 5.4 2.3 0.3 Same as case H except that
the ref. pi, fur calc. of hi
is changed to 32. MHDVBIBG.2UT
H 1.3 0.1. 0.5. Same as case H except that
The ref. pt. for calc. is
changed to 22. MEVB1G. 3UT
Notes: unit for variances is mmtt2, SB-Sampie
cire used in gstimating variances. HCV :
No.of Multi-conirol var. used. MDUBD??,??
refer to output file names. In case of
methods K & L, as the variations were
very small for the different cases, only
one representative value is listed. pl.
stands for plane.