PHOTOINTERPRETATION

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP79B00314A000900050001-7
• Release Decision: RIPPUB
• Original Classification: S
• Document Page Count: 10
• Document Creation Date: December 15, 2016
• Document Release Date: December 29, 2003
• Sequence Number: 1
• Publication Date: December 7, 1966
• Content Type: REPORT

Approved For Release 2004/02/12 : CIA-RDP79B00314A000900050001-7
IN'FORMATION REQUEST FLAG
THIS FORM IS USED TO SAFEGUARD FROM INADVERTENT DESTRUCTION RECORDS THAT ARE INVOLVED IN FREEDOM OF INFORMATION
ACT OR PRIVACY ACT REQUESTS, LITIGATION, OR MATTERS UNDER INVESTIGATION BY THE DEPARTMENT OF JUSTICE OR CONGRESS.
IT MUST BE FILED IN THE FRONT OF A FOLDER TO LIST ALL DOCUMENTS IN THE FOLDER RELATED TO A SPECIFIC REQUEST (one form
for each request), OR AS A SINGLE ATTACHMENT TO EACH INDIVIDUAL DOCUMENT. IT MUST NOT BE REMOVED FROM THE DOCUMENT OR
FOLDER UNTIL THE RECORD SERIES IS DUE FOR DESTRUCTION IN ACCORDANCE WITH APPROVED RECORDS CONTROL SCHEDULES AT
WHICH TIME DISPOSITION OF FLAGGED RECORDS WILL BE CHECKED WITH THE COMPONENT RECORDS MANAGEMENT OFFICER AND
OTHER AUTHORITIES AND WILL BE DEPENDENT ON THE STATUS OF EACH REQUEST.
TYPE OF REQUEST
PRIVACY ACT
CONGRESSIONAL INVESTIGATIONS
OTHER (specify)
FOIA
DEPT. OF JUSTICE INVESTIGATIONS
LITIGATION
MANDATORY CLASSIFICATION REVIEW
REQUESTER
REQUEST NO.
SUBJECT REQUEST
DATE OF REQUEST
. / / "
U
"
_
REQUEST . INCLUDE FILE NO., SYMBOL, DATE AND SUBJECT
DOCUMENTS SUBJECT
T
O
CONTINUED ON SEPARATE PAGE
REMARKS
CONTINUED ON SEPARATE PAGE
DATE OF COMPONENT RESPONSE DATE APPEALED (if applicable)
DO NOT REMOVE THIS FORM FROM THIS FILE
FORM 4016
9.77 Approved For Release 2004/02/12 : CIA-RDP79B00314A000900050001-7
 3". lerfg.T
Approved For Release x`4/02 RDP79B00314A000900050001-7
DEC
1. Aerial photography has become increasingly important to the
Agency both as a means of receiving initial information of military
and economic importance and of confirming information from other sources.
Recognition of this has stimulated the design of better collection
systems for receiving higher resolution photographs for processing.
Simultaneously, the entire process from initial collection through
interpretation at NPIC has gone through repeated reviews. It has
become apparent that increasing the number of high resolution photo-
graphs and the amount of information in each is not an unmixed blessing.
It has resulted in increased demands and consequent stresses on the
NPIC System.
2. Briefly, the system works as follows: The film is processed
and ancillary and historical data on specific targets in the photographs,
or on geographic coordinates are collated. These procedures are being
increasingly optimized from a production standpoint and more and more
automated with more or less concern with the primary user group - the
photointerpretation staffs.
3. Then the individual interpreter, whose perception is in part
predetermined by his prior experience and by the requirements levied on
him, abstracts some of the requirements levied on him, abstracts some
of the information in the photograph and communicates it primarily in
the form of reports to the analyst.
4. It is the thesis of this report that examination of the data
processing system at NPIC is best conducted in terms of the photointer-
preter. His is a key role. He is the primary mediator between the
potential information available in the photograph and the actual data
transmitted. It is our suggestion that we examine the interaction of
interpreter and photograph in the manner presented below. The data
so gained is needed to increase the interpreter's efficiency. It is
also needed to better coordinate the activities of the other components
of the system, both those which present the interpreter with the raw
information, and those analysts who need this data for further evalu-
ation.
5. Long Range Objectives:
If it is accepted that the NPIC information system may be
meaningfully analyzed in terms of the interpreter, then different
catagories of information not presently available are required. The
process whereby the interpreter both detects the pertinent information
in the photograph, interweaves informational items outside the immediate
photographs, and labels areas of the photograph ("SAM site" or "SUB",
etc.) must be quantitatively determined. Moreover, the process should
be studied in terms of the following factors:
G140U/
S1 logo Ulm
Approved For Release 2064/02/12 : CIA-RDP79fO63144000900050001-7
 Approved For Release 2004/02/1 4 R P79B00314A000900050001-7
(1) The interpreter's instructions set i.e., the intelligence
requirements list and his interpretation thereof.
(2) The constraints imposed by psychophysical and physio-
logical factors.
(3) The "image-quality" features of the actual photograph.
(4) The types of information added to material actually "seen"
in the photograph-..
(5) The number of individual photographs handled per unit time
and the "contextual" complexity of each.
6. The Photointerpretator's Task:
On its most elementary level a photograph may be considered as
a matrix of irregular dots, elements, which vary in size, shade of gray,
and clustering or organization. As the interpreter primed by his in-
structions to find representations of particular groups of objects looks
upon this matrix, he does essentially two things, first, he scans the
photograph and "targets in" on a particular subset of the matrix which
is sufficiently distinct in organization, or tonal patterning to be de-
tectable against all the other neighboring points which now constitute
its background. Secondly, he concentrates on this chosen subset by
enlarging it, by relating it to neighboring subsets, and by relating
it to other comparable subsets in previous photographs.
7. At some point, a label, a name, is assigned to the subset.
It this label is one within the instruction set the interpreter may
proceed further in "interpretion" - examining smaller groupings of
elements within the "object". The meaningfulness, the labeling, of some
of the smaller groupings and the rejection of others is governed, in
part, by the label assigned to the larger subset. Or the second process
may be reversed in that small clusters of elements act as cues for the
interpreter to look selectively at neighboring small groups. Varying
degrees of importance are assigned to the presence or absence of certain
clusters or patterns or tonal variations. They are intergrated into a
concept and labeled.
8. The targeting in and the final interpretation are not independent
processes, nor do they necessarily take place in sequence. We can, though,
for purposes of analysis isolate the two. Of the two, examination of the
first process, the initial scan and localization of the area of interest,
should be the first to be investigated. It is more dependent on basic
visual processes which can be studied psycho-physically. It permits us to
evaluate at the same time the factors of instruction set and image quality.
In addition, it is necessary as background information for a serious in-
vestigation of how the interpreter relates units into a concept.
Approved For Release 2004/02/
kDP79B00314A000900050001-7
 Approved For Release 2004/02/12 : CIA-RDP79B00314A000900050001-7
9. Experimental objectives:
To understand how the interpreter scans and articulates the
photo into figure and background, the primary question to be asked
experimentally is: What factor or combination of factors determine
that one or another subset of a photograph is detected and emphasized.
The fundamental factors, our independent variables, are: size, edge,
contrast, tonal patterning within a cluster (i.e., texture) connectivity
of elements in space or in tone, (i.e., shape), average intensity of
the subset, average intensity of the "background", tonal patterning of
the background, and the nature and informational content of competing
clusters as well as the separation between "objects".
10. Our conditioning variables are: the explicitness or non-
directiveness of the search instructions, the subject's familiarity
with the task, the time allowed for searching, and, most important,
the type of photograph. That is to say, a photograph of a sub in a basin,
descriptively a gray smudge on a slightly different shade of gray, differs
radically from a photograph of an industrial constellation with its
overabundance of shadings, borders, and distinctive features. There is
no a priori reason to expect that those fundamental visual parameters
most important in detection in one type of photograph will be of equal
importance or of any importance in photographs of a different type.
11. Our dependant variables will be, depending on the particular
experiment, the subject's response time,, or his accuracy. The importance
of the independent variables listed above has in varying degrees already
been documented. But they have been analyzed in simple patterns such as
homogenous light spots under conditions where the experimenter has
complete control of the independent variables.
12. But we cannot rig a stimulus so that only one or two of the
parameters of interest are present. We must begin with real photographs
and deduce what the relevant factors are. We must be able to elicit a
subject's response to the initial photograph, change rapidly one or more
factors, record the new response, distort a portion of the photograph or
blot out areas or relocate a subset spatially, and repeat the process
until we are experimentally certain that, for the given type of photo-
graph, we have uncovered those factors which optimize detection; or,
coversely, those factors which maximally degrade the
detection is possible.
image so
that no
13. For this purpose,
using
ordinary photographic means, we could
create ahead of time a large
file
of a photograph and all variations of
it that we may need. Aside
from
the amount of work involved in compiling
each file, this method would
not
work. First of all, even restricting
our instructions so that only a few objects were to be detected, the
numbers of photographs involved would soon constitute a retrieval
problem, since all possible variations would need to be available though,
depending on the subject's response, only a part of the file would be
utilized.
Approved For Release 2004/02/ 2:CFA-RDP79B00314A000900050001-7
 S T
Approved For Release 2004/02/1 RRDP79B00314A000900050001-7
14. For example, using a limited number of objects of interest,
a homogeneous background, and changing only one parameter at a time with
respect to the master photograph we would need as a minimum: Number of
photographs/file=(number of objects) x (number of size changes) x (number
of intensity changes) x (number of distance between object + neighbors)
x (number of edge accentuations) x (number of texture changes) + 1 (master
photograph). If we use three objects of interest, five gradations in
size, ten levels of intensity, five combinations of distances between
objects, four edge accentuations and five texture types, we would need
15000 photographic variations of the orignial. Using a patterned back-
ground, introducing noise, or the ability to distort or degrade only a
portion of one object, or to change two factors in one photograph, or to
rotate an object or to remove an object from its normal position and re-
locate it in another part of the photograph, would rapidly increase the
number of photographs in the file.
15. When we had completed a file which tests the effects of discrete,
relatively large changes in our key factors we would still need a way to
produce continuous, small changes in those variables which the subject's
responses define as crucial.
16. To produce adequate stimuli and record the subject's response
is to fulfill some of our requirements. In addition, we need to accu-
rately measure the amount of each factor in the master photograph and
in its variations if we are to generate and test hypotheses relating
critical factors to photograph type and instruction set.
17. Equipment Requirements:
We are saying, in brief, that in order to enhance the physical
qualities of a photograph usefully we need hard date, on which ones the
interpreter responds to under varying conditions. To obtain this data
efficiently we need certain specialized equipment. These hardware speci-
fications are implicit in the proposed experimentation detailed above.
Let us rephrase our requirements in terms of the type of equipment which
will give us the capability of truly interfacing man and machine.
1. The visual display presented to the interpreter must
satisfactorily simulate "real" photographs, or real
photograph degraded by noise. At other times, we
would need artificial photographs generated by patterns
of discrete dots which conform to particular mathe-
matical models.
2. The stimulus of the photograph and the response of
the subject can not be static or oneway. On the con-
trary, the man's response must initiate changes in
some qualities of the image which in turn will change
the man's response. Flexibility, ease, and rapidity
in changing parts of the pattern are prime considera-
tions.
Approved For Release 2004/02/12: 791300314A000900050001-7
 Approved For Release 2004/02/12 CIA-RDP79B00314A000900050001-7
3. A permanent record of the status of each element in
the photograph and of the subsets articulated by the
interpreter for each change in the photograph must
be available during on-going data collection or post
hoc data analysis.
18. Present computer design capability would argue for the desira-
bility of an analogue-type visual presentation to the photointerpretor-
subject. In addition, the subject must be able to indicate directly to
the machine his area of attention by simply pointing with a lite-pen,
curser, etc. The experimenter must be able to direct real-time changes
in the display, especially in those areas of indicated primary attention.
Because of the precision required in such partial picture substitution
and syntheses, it would appear desirable that digital processing of such
areas take place. Representation of such digitally-processed areas on the
visual display would require intermittent digital to analogue conversion
with subsequent replacement. In order to approach the ideal of on-line
processing, it would seem expedient that the system store a fully digit-
ized version of the analogue presentation, which of course could be pre-
pared off-line in advance of the individual experiment.
19. The foregoing criteria are for a hybrid (analogue plus
digital) system for display and manipulation of images; a system which
is not at present existant but one which does not demand capabilities
too far beyond the present state of the art. A preliminary plan of
such a system would be as follows: (See diagram)
Approved For Release 2004/02/12 CIA-4DP79B00314A000900050001-7
 Approved For Release 2004102P12 : CI 900314A000900050001-7
DISPLAY SYSTEM
0
Scanner
A/D
Inserting
original
photograph
I* I
timer E. 1 r..._.~
Photograph
fixed
disc
for
remote
memory
storage
Storage
tube
D/A
two
tape
units]
inter-
changehble
disc
programmer's
equipment
Computor and controls
for storage of info
and for revising image
T
facsimile'
subject
CRT
Misc.
controls
tote pen
or
cursor
Approved For Release 2004/02/12 : 4HOU91300314A000900050001-7
 Approved For Release 2004/42/12 : R 9B00314A000900050001-7
Insertion of Original Photograph:
We are assuming a 8" x 10" photograph or transparency and a scanner
with a spot size it in lateral resolution or 2.5 x 10;5 inches in area.
The photograph then has 8 x 10/2.5 x 1005 or 32 x 105 such elements. If
we assume that 32 levels or 25 levels of gray will be required, than an
additional 4 bits of information must be stored for each element making
a total number of bits of 4 x 25 (=32) x 105=27 x 105 = 128 x 107 bits to
be stored in the remote memory disc and tagged so that portions of the
image are readily retrivable.
Front-end:
The displays to the subject includes a facsiLmile system in addition
to the usual TV CRT display. Where the TV display is usually capable of
1000 lines resolution and about 7-10 levels of gray, a fascimile system
can resolive 2000 lines/1" and reproduce 64 shades of gray. The CRT
display with less potential information is relatively easier for program
and would be preferable in the production model. However, until we can
spacify which qualitative analysis are valid with the resolution-and gray
scale-limited CRT, both displays are required. The CRT for the operator
is needed primarily for object location. The miscellaneous controls are
to be used to call forth programs from the digital computor for changing
portions of the display and for gross variations of contrast or texture
by analog means.
It should be pointed out here that one unknown in the system is
which changes can be controlled from analogue devices included in the
miscellaneous controls without reducing the efficiency of the entire
system.
Computor and Controls:
The computor must be capable of storing information about the entire
image (the remote memory storage), of storaging programs for manipulating
changes in size, texture, etc. in parts of the image (interchangeable disc
and tapes), and of controlling the revised image or the display via the
storage tube and associated electronic controls.
Of the 128 x l06bits storing the original image, let us assume we
will be namipulating a 1" x 1" portion of the picture at any one ime.
The computer then may, at any time be processing l~ = X x 10 or
160,000 bits of information at a time. By itself, it is reasonable to
expect that this amount of information can indeed be processed in real
time (i.e. within 30 sec to 1 min.) However, within that time, the computor
must retrieve the information about that particular section of the photo-
graph, run it through one or more programs, (eg, change size of object A,
make all densities below D1 equal to D1, and transfer object A to position
(X1 + H) (Y1 + L) on the screen) and revise the screen image.
Approved For Release 2004102/12 : f79B00314A000900050001-7
 Approved For Release 2004/02/12 I G*T79B00314A000900050001-7
Since factors in estimating time requirements include the programming
subroutines needed and the physical characteristics of the images to be
manipulated, whether manipulation of more than one change at a time is
possible in real time has not yet been answered.
We have assumed as our minimum computor requirements a CDC 1700, a
PDP 7, or possibly a SDC 930. That is a 32,000 word (16 bit/word) machine.
Between the computor and the display a storage tube and related
electronics has been inserted. This is because the output rate of present-
day computors or of computors in the forseable future is well under a
megacycle, while TV imput requires a 4 megacycle rate.
Programmers Section:
The card, reader, etc, are not strictly part of the operating system,
bur are for use of the programmer in writing and debugging the computor
programs for inducing the changes in the image.
Implementing the System:
Preliminary talks with representations of the
indicate that the proposed system can be built from presently
available components in 9 months to 1 year. The estimated cost, exclusive
of programming would be about for the computor and
for the other components. If the system is to succeed, three programs wou
need to be begun simultaneously; 1) The building of the display system,
2) The development of the software package, 3) The strategy of parameter
manipulation in an unconventional experimental design.
The personnel requirement would be; 1) An electronic engineer,
conversant with TV display systems and computors, to write the explicit
specifications for the system, as well as to maintain it in use, 2) A
full-time programmer with experience in machine programming for pictorial
displays, 3) An experimental psychologist to formulate the experimental
strategy and runs the actual expermentation.
It is suggested that we develop the display system by outside contract,
while the programming and actual experimentation be done in-house.
Approved For Release 2004/02/12 :
25X1
25X1
 Approved For Release 2004/02A146,14 DP79B00314A000900050001-7
Object: To evaluate the factors in an aereal photograph (variations in
contrast, texture, size, border, location of an object relative
to its background and number of competing objects) taking the
parameters, 1 to n at a time, which increase the probability that
a photointerpreter will detect an object, or conversely maximally
minimize the probability of detection.
Method: Given a particular type of aereal photograph and a particular
set of instructions, to record the subject's response time and
object chosen, and on the basis of his response, iteratively
manipulate parameters of the image, until the combination of
parameters yielding optimum detection have been determined.
Equipment Needed: A CRT and facsmile display of the image under control
of a digital computor whereby the operator can implement local
changes in border, size, etc., in the display in real time in
response to the subject's response.
Approved For Release 2004/02/12: CIA-RD
F 3147~'ffb