COMMENTS ON PROJECT 6607 REPORT

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP79B00873A000100010086-9
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 4
• Document Creation Date: December 28, 2016
• Document Release Date: November 1, 2012
• Sequence Number: 86
• Publication Date: March 12, 1969
• Content Type: MEMO

Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9 (1
March 12f 19,69
TECHNICAL MEMORANDUM
TO;
FROM
SUBJECT;
REFERENCE;
Comments on Project 6607 Report
300301'TM-69,E-2 25X1
1. I do not believe that the statement of objectives (Introduction)
are sufficient. In particular, emphasis should be placed on the
determination of photographically related error.. I do not believe
anyone would agree that the program was all-encompassing enough
to answer?que'stion number 1. Question 3 has too many human
factors overtones.? It should be re-phrased. Question 4 was
not an objective. The intent by changing resolution was to
see if the basic conclusions were altered by changing the system
performance in a known way. I am not sure question 5 was really
an objective.
'2. It would have helped if the pages had been numbered, even if
by hand.
3. . I still don't understand why they changed from 3 to 2 operators.
I calculate that this should have changed the confidence interval
by a factor of .
Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9
 Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9
25X1
or about 40%. I guess in most cases, this would not have overlapped
the systematic error, so it really doesn't,make much difference.
I withdraw the comment. I would have liked to see a little more
exploration of errors and confidence'limits as a function of
density with the sensitometric data superimposed.
Wording is not good. in general. Page 2 of Section II is a good
example. I found their definition of error a little confusing,.
-After a while, I figured. out that'they generally referred to the
random (precision-type) of error as "error". In this case,
interaction of operator. and photographic effects become lumped
in the systematic error which is a dynamic type of error. The.
conclusion then is that a systematic error is introduced by the
nature of the'photography.
5. On page 2`of Section III, I do not feel that determining the
existence-Of a "positive correlation" is a proper-expression of
objective. A much 'stronger relationship would be required to
show'that the operator was materially affected by the image structure
and such a relationship was, in fact, shown.;
6. It really wasn't necessary to make the h`_(P) = h (-pr .assumption
in the microdensitometer section where the spread function
correction was calculated. I think the recursive solution will
still work.
Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9
 Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9
25X1
.10
7. I don't disagree with any of your comments.
8.. As a general conclusion, I-judge that this work is very
.significant. It shows that the situation-induced error is large
compared to the precision involved in making repeated measure-
ments by comparators in the presence of grain noise.
9. I would have liked to have seen an analysis for the three
operators showing paired differences and an analysis of bias,
if any, between operators,.to support their arguments on
confidence limits.
10. Figure,9, T think, shows a lower deviation for 8 than for 12
readings (I think, anyway; my copy is bad.). What were the
confidence limits on.a.
11. I wish.-.they had used a more consistent verbal notation throughout.
I,got confused on some of the errors. Film random error was
sometimes called deviation, for example.
12. The correction procedure is.very clever and supports my earlier
contention about the impact of image restoration on measurement.
13. I didn't care much for the format-and content of the summary and
conclusions section. It needs-to be less wordy, more punchy
.and not based on the questions (which I think should be re-
worded anyway).
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9
 Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9
25X1
14. I do not think that dual gamma will affect these conclusions
much for low contrast objects, and most small objects will
fall into this category. If one can get some estimate of the
spread function in the same general density and contrast range,
correction could probably me made. For higher contrasts,
.problems will arise. I believe that a similar experiment
should be conducted on dual gamma; using relatively low
achieved'AD?s, say in the 0.5 range.
The?basic,.phenomenon which causes the systematic error seems
to involve the detailed"image structure. Since-we could not
cover all cases of image structure anyway, the addition of
dual gamma simply increases the size of an already very large
matrix of possible responses. Consequently, ,I believe that
dual gamma may change the conclusion about correction (at
it
least in cases with larger AD or wherein the choice of spread
function-estimator is limited)-. The conclusion regarding an
interaction between structure and measurement should not be
affected. This is the fundamental result and is of the most
importance to the mensuration field.
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/11/01: CIA-RDP79B00873A000100010086-9