. 
54 
SYMPOSIUM PHOTO INTERPRETATION, DELFT 1962 
control and disarmament plan. It is highly significant that during the Arms 
Control discussions in Geneva in 1958, the representatives of the United 
States of America and the representatives of the Union of Soviet Socialist 
Republics both proposed that aerial reconnaissance of one sort or another 
should be an important tool in support of an arms inspection plan [4]. Table 2 
suggests the scope of aerial observation in support of such programs. It is 
difficult to estimate how much actual photography would result from such 
an activity, but an estimate of 100,000 prints per week seems at the same time 
possible and almost unbelievable [5, 6]. 
Table 2. Anticipated scope of aerial observation in support of arms control * I 
A. 
B. 
C. 
D. 
E. 
F. 
G. 
Missiles 
Long range aircraft 
Tactical aircraft 
Ground forces 
Missile-launching submarines 
Other naval forces 
Other to include: 
Industry 
T ranspor tation 
R & D Centers 
1. Thousands of individual observation sites. 
2. Tens or hundreds of thousands of photos generated 
on each complete “look”. 
3. Frequency of looks undetermined but probably, to 
include “continuous sampling”. 
With respect to both weather and arms control reconnaissance, it appears 
that we can well collect photography required to do these jobs, but the ques 
tion remains as to how we can process and efficiently use all of the resultant 
material. 
At the System Development Corporation, we are conducting research 
programs which bear on the specific areas I have pointed out. They are areas 
in which computers can do a specific, significant, and unique job of system 
support. We have tackled the job of automatic photo screening, and are also 
involved in a look at the nature and function of the photo data base, more 
specifically, at the problems of man-machine communication in free text. 
I would like, now, to give you a rapid summary of our research in each of 
these areas. 
S.D.C. research on computer applications in photo interpretation 
Pattern recognition 
The concept of automatic pattern recognition is, of course, not new. It has 
been studied extensively with respect to machine recognition of printed type 
and hand written material [7]. To a much more limited extent efforts have 
been made in the direction of a system to read, i.e., screen, photographs [8, 9, 
10, 11]. Table 3 summarizes the problems and the prospects with respect to 
such devices.