PROPOSAL FOR AUTOMATIC IMAGE RECOGNITION BY COHERENT OPTICAL TECHNIQUES

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78B04770A002300020012-3
• Release Decision: RIPPUB
• Original Classification: C
• Document Page Count: 9
• Document Creation Date: December 28, 2016
• Document Release Date: March 30, 2005
• Sequence Number: 12
• Publication Date: August 28, 1964
• Content Type: LETTER

Approved For Release 2005/05/02 : CIA-RDP78B04770A00230002001i3&.
CONFIDENTIAL
Declass Review by NGA.
Subject: PROPOSAL FOR AUTOMATIC IMAGE RECOGNITION
BY COHERENT OPTICAL TECHNIQUES
144-20-9
Copy / of
25X1
Attached is a Technical Discussion of the work program which we
are proposing on the subject above. From the standpoint of
content, I believe it is responsive to the items we have discussed.
From the standpoint of completeness, I believe it is a good first
cut at the major questions that we have current knowledge of.
Successful completion of the proposed work program will not answer
all of the questions that any of us can pose. Rather, it is a
look at the most important real-world problem areas. Together
with this first try, we are proposing work which will enable us
to get understanding and appreciation of the nature of the process
as it relates to your end use. We will not manipulate all of the
conceivable variables in all of their possible permutations and
combinations. The expense of such an exhaustive experiment is
not warranted at this time. We will gain an appreciation which
will permit you and your associates to assess the potential of
the concept and be able to make a hard-headed judgment as it
relates to your next step.
The estimated costs for this program are described in the attached
Exhibit A. The option described as Part 2 is that relating to the
use of the image orthicon camera chain. The pricing which is
shown for Part 2 assumes that the image orthicon chain which is
in-house will be furnished free-of-charge to your program. I am
hopeful that the experimental results detailed in the Appendix
of the attached Technical Discussion will provide the graphic
evidence to result in a favorable decision on the inclusion of
Part 2.
The output of our work will be reports as described in the attach-
ment. The final report will be delivered thirty (30) weeks after
contract authorization is received. We would also assume that
this work, if authorized, would be a task under our basic agree-
ment.
Sincerely,
Enclosure
0p piRaff e 261
THI6 1R47=91.PL i' :: W AS IKi:S?'A1IA}t AFFECTD ThE
GS 7119 HFJIYffl LTATFS CUM THE
25X1
4AO02300020012-3 GROUP ;
EXCLUDED FROM AUTOMATIC
 Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
Next 16 Page(s) In Document Exempt
Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
 Approved For Release 2005/05/02 : CIA-RDP78B04770A002300020012-3
25 August 1964
1tesearch Ob ective
etic Imp a Rec ition b Coherent 0 ical Tecues
1. I xtro d.u.at;
Research is proposed to investigate the feasibility of utilizing
coherent light techniques to automatically or semiautomatically recognize
photo images.
This investigation, would be one of several being performed by
the intelligence community to evaluate various techniques which may
be applicable to the development of automated devices which would
aid the interpreter in performing his image recognition tasks. Many
of these techniques under investigation are primarily linear processes
based on the sequential e3mud,netton ( by an electro-optical system) of
each bit of data in the total image and the subsequent c npsris of these
bits to a. previously learned prototype image of a target. Foremost of
this type of image processing are the adaptive-memory devices such as the
Perceptron and other similar computers. Although these types of devices
can auto ticslly identify targets, they seem to have several serious
deficiencies:.
1.1 Linear processing techniques are very bandwidth-
limited -- e.g., the time of processing per unit of image
area per image-packing density is very long. In other words,
an electro--optical, adaptive-computer system requires more time
to identify an image than jades a poor or mediocre human inter-
preter. A human interpreter does not examine an image bit by
bit but by integration of many hundreds or thousands of image
bits simultaneously.
1.2 Adoptive memory techniques have shown little
capability to :generalize" from a learned prototype image.
They are unable to recognize an image which differs appreci-
ably from the learned prototype due to shadow, perspective,
and incomplete or distorted image detail.
It is primarily be{:a.use of these two deficiencies that it is necessary
to investigate cozipletely different concept which might be applied to
image recognition.
Approved For Release 2005/05/02: CIA-RDP78B04770A002360020012-3
 Approved For Release 200$/05102 : CIA-R0R78B04770A002300020012-3
Recent research has demonstrated that certain types of optical
filters for apecific images will permit recognition of only that image
when it is viewed in a scene containing other "unwanted." objects. 'hi:
type of filter is esseutial.3;r a ditfractin image of the object, more
etonly called a hologram. These types of spatial filters, or holo-
grams, permit the entire image to be proceased" simultenemisly;
therefore, the speed of recognition would be reduced considerably over
that of linear recognition systems. There is also some indication from
laboratory work that recognition with these filters is little affected
by perspective and incomplete Image detail,
Although the basic concepts of spatial filtering have been demon-
strated., there are many questions ramming that must be solved before
the technique can be fully evaluated and related to a possible auto-
matic recognition system. The purpose of the proposed research is to
investigate these more salient unresolved problems.
2. ~CMCFJLt
2.1 Purpose
Although there are a number of questions relative to
the Implementation and operational, use of holograms in a recogni-
tion system, the basic purpose of the proposed research is to
acquire more knowledge about structure and variations in the
itself and its limitations in image recognition tech-
Lues. This additional knowledge should allow a more ccanpre-
benaive evaluation of the potentials of holograms and provide
insight into the feasibility of further development of a
recognition system based an holograms.
2.2 Scope
This project would incorporate, essentially, two
levels of experimentation:
2.2.1 Investi ;ati.on of the interrelationships
of hologram structure and image variables.
2.2.2 Investigation of probleiiis relating to the
optical system and the recording medium.
Rg,,rents
3.1 This project would inclu: a optical experimentation and
analysis leading to answers for at least the following unresolved
questions :
Approved For Release 2005/05/02 CIA-RDP78BO477OA002300020012-3
 Approved For Release 2005/05/02: CIA-RDP78BO477OA002300020012-3
..jar =E
3.1.1 Now obscure can a target be and still be
recognized? e.g., low contrast targets and partially
hidden target s .
3.1.2 How much information can a hologram, contain
before the desired target information is "lost"? e.g.,
how much of a photograph can be searched in the field of
view?
3.1.3 What are the effects of shadows?
3.1.4 What are the effects of image scale?
3.1.5 What are the effects of target perspective?
3.1.6 What are the effects of image packing density'?
3.1.7 What are the effects of target orientation?
.3.1.8 What are the effects of multiple, identical
targets in the same field of view?
3.1.9 How many different types of targets can one
hologram accommodate?
3.2 There are probably other image variables which may affect
the basic structure of holograms. It is expected that these will be
investigated when they come to light.
3.3 In addition to these basic questions of target variables
and their effect upon holograms, there are other related problems
for which investigation is desired. These are primarily concerned
with the optical system, filters and the recording medium.. These
problems include, but are not limited to:
3.3.1 How do glass plates affect the optical path?
3.3.2 Are liquid gates necessary?
3.3.3 How seriously does vibration affect the
optical system performance?
3.3.4 What are the potentials and limitations of
recording an photoplastic film?
Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
 Approved For Release 2005/05/02: CIA-RDP7.8.04770A002300020012-3
3.3.5 Sow precisely must the filter be located and
aligned in the optical system?
5.3.6 What are the relative merits of phase and
amplitude filters?
3.3.7 What are the expected resolution limits of
the hologram system?
3.3.8 What are the predicted time conetents assoc-
iated with the sceling and orientation variables?
3.3.9 Wbot is the optimum modulation transfer
function of such a system?
3.3.10 What is the optimum signal to noise ratio?
3.3.11 What are the relationships between system
processing time, working area, and the resolution of the
photograph?
3.3.12 What is the feasibility and complexity of
real-time filter generation?
3.3,13 Whet are the beet techniques for reducing
amplitude of the reference beam of a two-beam interfere-
meter without changing the characteristics of the inter-
ference pattern in order to achieve interference in the
desired order.
t ~`
Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
 25X1 cJ
Approved For Release 2005/05/02: CIA-RDP7
25X1
25X1
25X1
vet ~ u~~ ~ .
25X1
Approved For Release 2005/05/02 CIA-RDP78BO477OA002300020012-3
 F
Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
SECRET
(When Filled In)
TECHNICAL BACKGROUND PROCUREMENT INFORMATION
I. Contractor
A. Name and address:
25X1
B. Evaluation of previous performance:
a concurrent contract.
Excellent work being performed on
II. Brief description of this procurement: A study project involving
Automatic Image Recognition by Coherent Optical Techniques.
A. Deliverable items: Reports only
B. Is this procurement for other than a stahda-rd, "off the shelf" or slightly
modified commercial item? If "yes", is it anticipated that
any more of this unit will be procured? If so, a complete
set of directly reproducible manufacturing drawings and specifications
would normallybe included in this procurement. Comments:
C. Will contract cover a period of more than 90 days ? yes
If "yes", are progress reports desired? yes If so, indicate fre-
quency, content and number of copies desired:
Monthly informal progress reports (3 copies)
D. Is any Government-owned property to be provided to the contractor?
If so, list and indicate its availability (where, when,
etc.) Eutony Image Orthocon System S?rill be utilized in this study, This
SECRET
(When Filled In)
Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
25X1
25X1
 Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3
SECRET
(When Filled In)
25X1
25X1
system was developed by ~~or D/DS&T/ORD. However, it will remain at
for a time. NPIC has permission from ORD to utilize this equipment
for this project.
X1
2. The specifications and/or drawings are
Unclassified
4. Contractor personnel known to be aware of this proposed procurement:
5. Other security information
has a basic contract with the agency.
L 25X1
III. Reasons for selection of this source. If other` sources were considered, indi-
cate results. If no other sources were considered, list the reasons why this
firm is considered to be uniquely qualified to perform this work.
25X4,,;
has developed certain techniques and information which are considered
proprietory. In addition, they are the most advanced group in the applications
of this overall concept.
25X1
IV. Technical contact
In the event additional space is required, use the reverse side(s) of this form, with
a reference to the item number to which the comment applies.
SECRET
(When Filled In)
E. Is any special tooling involved? Np _
F. Security:
1. Association with the Sponsor is
Approved For Release 2005/05/02 : CIA-RDP78BO477OA002300020012-3