stinguish flaw 
rpose. 
of TOHOKU 
. 7. are Picture 
at field test at 
ed image data. 
polished part 
; And also 
| the image of 
this case, the 
te strength. 
dition is easily 
ht condition of 
d data. 
d Value 
nalysis (mm 
| Too Dark 
2 test1 | test2 
      
   
58 
57 
56 
55 
58 
39 
4| 22 
31] 21 
3] 18 
$21 17 
0] 15 
ei] 14 
| Far Side 
km / hour 
ated 
ted 
ated 
urface with the 
letect from the 
camera. 
'essing time on 
kes about Five 
gth. ( Approxi- 
performance of 
it this system 1$ 
system, that an 
natic inspection 
acks. 
  
ACCURACY IMPROVEMENT IN COMPUTATIONAL CLOSE-RANGE PHOTOGRAMMETRY BY THE USE OF MULTI-CONTROL VARIABLE {NCV) 
MONTE CARLO METHOD 
H.N.Nagaraja, Assoc Professor, University of Zimbabwe, Zimbabwe 
Commission V, worhing group V/7 
KEY WORDS: 
ABSTRACT 
Statistics, Application, Simulation, Error, Accuracy, Ciose-Menge, Method, 
Accuracy improvement hy use of multi-conirol variate method MCV} among other Monte Carlo methode is presented, 
The INCV) method is particularly applicable in close-range photogrammetry as the accuracy of computed three 
dimensional coordinates of any detail point can be improved by the sisultanecus comparison and rontideralion of 
à set of all available object space control point coordinates. 
1. INTRODUCTION 
Monte Carlo techniques are often used in many 
scientific and other disciplines and an isportant 
part nf this effort is direried towards variance 
reduction in systems analyses, The applicability of 
some of these techniques to the normal case of 
terrestrial and close-range photogramseiry has been 
identified, developed and partly demonstrated in 
iNagaraja,1971,92), Though at first sicht it might 
seen that sampling procedures only apply in case of 
simulation studies, further reflection should 
indicate that conceptually and practically, it should 
be possible to incorporate these ideas in reduction 
of variante while dealing with practical data. 
However, this paper deals with à case study involving 
some data sets obtained by simulation only, yet 
closely conforming to actual data. It was not 
possible to use live or field dats in this study, 
The subject of variance reduction has received, 
outside photogrammetry, considerable attention and a 
number nf methods have been developed. Hence, there 
ère à few techniques that help to intreasp accuracy 
any hence efficiency of simulations,  — sometimes 
substantially, by producing less variable 
übservations, Accordingly, there is a need to study 
such sethods and possibilities, 
Applications of tigulation are net ne in 
photograsmetry — hut new applications are still 
possible. Variance reduction techniques seek either 
increase in precision {decreased variance) for a 
399 
fixed sample size of m Decresse in sample Size 
required to oblain a given degree of precision, 
Several authors have cautioned us in using these 
techniques without first ensuring their applicability 
and effectiveness, 1f properly used, these 
technigues ran provige tremendous increase in the 
efficiency of the sodel; however, if the intuition is 
faulty and the analyst does mel use a reasonable 
design, the technique can also be quite unpredictable 
and perhaps actually increase variante for 5pae 
techniques, Because of this charatterislit reason, a 
systematic and thorough study of the selected method 
is both desirable snd essential in adapting it for 
any specific application in photograsmetry, 
The multi-control varisbies technique applies very 
weil when there tiose repetition or a near 
equivalent to the process we arp using in simulating 
that can be treated theoretically, Thus, in the 
normal case of clpse-range  phologramsetry, Just as 
alep in aerial pholograssetry, we have an equivalence 
between cosputation of three disencional coordinates 
of 8 non-controi object space point and that of à 
given or knows control object Space point. We can 
then simulate the least acturate Y-Coordinates isay, 
when it is known that the Y-coordinate axis and 
Camera — axis are parsllell cof a selected object 
space point and that of the known control point 
sisultanenusly, using sa&bP random nusber Stresss in 
both computations, The difference in the known and 
computed coordinate of the contro! object space 
point is indeed an estimate nf the correction thet 
tan be conveniently applied to the computes 
coordinate nf the selected non-control chiert space 
PE 8 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996