AUTOMATIC IMAGE REGISTRATION EXPERIMENTAL PROGRAM

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78B04770A000200010039-8
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 56
• Document Creation Date: December 28, 2016
• Document Release Date: July 1, 2004
• Sequence Number: 39
• Publication Date: October 1, 1967
• Content Type: REPORT

AUTOMATIC IMAGE REGISTRATION
EXPERIMENTAL PROGRAM
Declass Review by NGA.
For Release 2004/07/29 : CIA-RDP78B04770A000200010039-8
STAT
STAT
 STAT
Approved or Release 2004/07/29: CIA-RDP78B04770A000200010039-8
AUTOMATIC IMAGE REGISTRATION
EXPERIMENTAL PROGRAM
STAT
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
TABLE OF CONTENTS
1. Introduction
2. Technical Discussion
2.1 EROS Preparation
2.2 Experimental Program Description
2.2.1 Photo Materials Preparation
2.2.2 System Reference Data
2.2.3 Precision and "Zero Parallax" Determination
2.2.4 Clearance Error Measurements
2.2.5 Experimental Results
2.3 Image Point Selections
2.4 Data Evaluation
2.4.1 Measuring Mark vs. Raster Relationship
2.4.2 Raster Size Determination
2.4.3 Measuring Mark Motion Calibration
2.4.5 Stereo Parallax Errors
3. Conclusions and Recommendations
4. Appendix A - Technical Report - EROS
Appendix B - Film Diapositives - Scale 1:10,000
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
r
s
1
1.0 INTRODUCTION
All present automatic image registration systems are dependent for
operation upon the scanning of appreciable areas of the involved stereo images.
Additionally, current systems are limited to zero and first-order transformation
capabilities which correspond to translational and slope-correctional facilities.
Consequently, since usual terrain forms are not planar in nature, the registration
system achieves the best average registration over the scan area. This leads
naturally to the existence of registrational error at the central point of the
scanning pattern which is the point of mensuration in all systems. Little if any
test data has been available with which to evaluate the magnitude of averaging
errors when measuring with an automatic image registration system. The completion
of EROS which is described in the report attached as Appendix A, has presented
the first opportunity to derive test data which allows a useful evaluation of
this problem. This report describes the experimental program conducted for the
extraction of data, the analysis of that data, and the conclusions and consequent
recommendations derived therefrom.
2.0 TECHNICAL DISCUSSION'
This program has been executed in three parts:
- EROS preparation
- Experimentation and data recording
- Data evaluation
The essential components of these parts are discussed in the following paragraphs.
2.1 EROS Preparation
EROS in its original configuration was not supplied with any form of
measuring marks or attendant mensuration capability. Consequently, it was neces-
sary to incorporate these means into the optical system before tests were possible.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Realizing that EROS registers images in two channels with respect to
a single scanning raster, that the optical axes in EROS cannot be expected to
remain stationary and that it is impractical to introduce mensuration means
into the EROS transport system, it was decided to introduce the required measuring
marks at the real image planes within the eyepieces. Figure 2.1 illustrates
the optical-mechanical realization of this scheme. It should be noted with
respect to Figure 2.1 that the left measuring mark is stationary and the right
is moveable over a small motion range. The left mark then, has a fixed relation-
ship to the scanning pattern if one neglects accidental shifts of the pattern
owing to stray magnetics. The range of motion of the right mark need only be
small owing to the fact that any gross parallax between images is automatically
eliminated by the registration system. The position of the right mark is signified
by readings (0.0001 inch least reading) on micrometer barrels; one each for px
and py. The marks themselves are 0.0010 inch diameter apertures illuminated from
the rear by lamp sources. To allow the operator to clear parallaxes, drive rods
extend from the micrometer screws forward to the front of the instrument. Owing
to the fact that EROS's zoom optics slightly change the location in space of the
final real image when the zoom is changed, the additional required features were
focus adjustment mechanisms. These features merely allow the maintenance of
coplanarity of the photo images and the measuring marks-,this making the measuring
marks sharp under all zoom conditions.
2.2 Experimental Program Description
The experimental program conducted was comprised of several preparation
and data gathering operations. Figure 2.2 illustrates the composition of the pro-
gram.
1
J
t
t
r
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8 I
 MOMON ~ go was" Om ON an
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Movable Measuring Mark
t Stereo Image
Transformed
I............_~........
Fixed Measuring Mark
Transformed Measuring Mark
Left Stereo Image Measuring
- /rr+~~*~%?? > ? 4s and Image Mixed
of Measuring Mark 1
or
Virtual Location,
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Determine Raster
Center vs. left
Measuring Mark
(min/max zoom)
O
Bar Pattern Plate
0
Open Loop
Precision Tests
(min/max zoom
Determine Raster
Size at Film
(min/max zoom)
O
Closed Loop
Precision Tests
(min/max zoom)
Focal Length - 6 inches
Altitudes - 5, 10, and 15 thousand feet
Attitude - vertical
E/H Ratio - 0.6
Closed Loop
Clearance Error
Measurements
(min/max zoom)
"Zero"
Readings
Closed Loop
Accuracy Tests
(min/max zoom)
(Automatic vs. Manual
Parallax Clearance
Errors)
"Zero" Parallax Readings
w >~ a- m r 60-m mom -
Calibrate Meas.
Mark Motion at
Film
(min/max zoom)
Experimental Results
Motion
Calibration
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 s
t
I.
t
r
'I
c
L
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Referring to Figure 2.2, it is seen that the experimental program is
divided into five general categories of work, each of which is composed of
one or more individual elements. The first category of work relates to the
preparation of photographic material necessary to the completion of the
experimental program. The second category of work relates to the determina-
tion of certain EROS parameters which are crucial to the experimental program.
The third category of work relates to the establishment of precision levels
attainable for parallax clearances under differing circumstances and to the
parallax readings that correspond to zero parallax between the test images.
The fourth category relates to the determination of parallax clearance errors
associated with the automatic registration of stereo images of specific types
of object-space, terrain and feature shapes. Then in category five, the
numerical results of categories two, three, and four work must be brought
together in order to derive understandable results from the experimental
program. More detailed discussions of the categories and elements therein
are as follows.
2.2.1 Photo Materials Preparation
There are two reasons for photo materials preparation. First, it
is necessary to provide the stereo images of certain terrain configurations
containing likely causes of correlator error so that the effect of these
causes may be evaluated. Second, the photographic materials must be so
constituted as to provide control for the experimental work so that instrument-
induced errors not associated directly with correlation may be eliminated from
consideration. Three types of photographic imagery are necessary to achieve
these dual ends.
The first photographic image produced was that of an equal bar and space
rectangular pattern. The widths of the bars and spaces are 1/32". The spacing
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
of the bars and spaces has been used to calibrate the relationship between meas-
uring mark motion in the input image and difference between micrometer readings.
Additionally, this pattern has been used to ascertain the relationship between
the stationary measuring marks and the scanning pattern. To do this the corner
of the bar pattern is brought to the stationary measuring mark. Subsequently,
the video information produced by the scanning system was displayed in a coinci-
dent scan pattern on an oscilloscope. The resulting display was photographically
recorded and the image measured to determine the location of the corner of the
bar pattern within the framework of the scanning pattern. The objective of this
measurement was to determine the proximity of the measuring mark to the center
of the scanning pattern.
I
1
IpPlilbomwo This location is of importance in that when registering
the stereo image of curved terrain, the maximum parallax error occurs at the
center of the scanning pattern when the correlator transformation system has only
first order capability.
The second set of photographic materials produced were stereo images
exposed over the same terrain, but from three different altitudes; five, ten,
STAT and fifteen thousand feet using a
six-inch focal length RC5A camera (See
Appendix B for duplicate samples of film diapositives forming a stereopair taken
at five thousand feet altitude, scale 1:10,000.) The reason for using three
photographic scales is that one would like to ascertain the significance of
registration errors over the same object points under differing photographic
scale conditions as well as under differing zoom conditions.
In order to provide a reference for parallax readings made on stereo
t
I
I
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 a
s
1
J
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
photographs, it was necessary to prepare a third set of photographic materials.
These photographs are identical copies of the left-hand stereo images. They
were printed however, so that the area of the image to be used for the stereo
test is at the left side of the diapositive plate and in approximate coincidence
with the location of that same area on the right-hand member of the stereo pair.
These materials were produced in this manner for two reasons. First, if we
measure numerous points in identical photographic plates, then we should achieve
the same parallax reading for all points. This parallax reatling then is the
reference "zero parallax" reading. If, subsequently, we register two stereo
images and do not get the same parallax reading when the measuring mark is put
on the terrain, then the difference is the error in the automatic registration
process.
The reason for producing the above described identical images on the
left-hand side of the diapositive plate is that such a precaution materially
reduces any possible parallax error which might be found owing to the non-linear-
ity of transport motion in EROS. In essence, we want to determine the zero
parallax reading with the transport plate carriers in the same relative positions
as will be the case when we test using stereo inputs. It should be noted in this
context that the selection of points to be tested for stereo registration errors
were selected with consideration for the equality of total parallax between
points. Again, this was done to limit in so far as possible any changes in
relative position of the two transport photo carriers.
2.2.2 System Reference Data
As mentioned previously in the context of photo materials preparation,
the category of work relating to system reference data is included in the
experimental plan in order to relate EROS characteristics to the tests being
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
performed. In essence, we must know the relationship between the stationary (left)
measuring mark and the scanning raster. Similarly if we want to know the relation-
ship of the scanning raster to the measuring mark, we also need to know the area
represented in the film by the scanning raster during operation under various zoom
conditions. This is necessary in order that we may evaluate any parallax errors
in relationship to the height anomalies existing within the scanned area. Addi-
tionally, we must define the relationship between the differences in micrometer
readings and the motion of the measuring mark in the plane of the input photographs.
Through this calibration we can convert micrometer readings into microns in the
film which is necessary in order to derive the height errors in the terrain cor-
responding to these parallax errors.
Each of the above factors has been investigated at each of the two zoom
settings that have been used in the actual image mensuration tasks.
2.2.3 Precision and "Zero Parallax" Determination
As background to any stereo registration tasks, it is necessary to
establish the precision levels to which parallax may be cleared under ideal condi-
tions. Also, it is necessary to establish "zero parallax" readings, or in other
words, the micrometer position readings at which the two measuring marks are in
virtual superposition on the scanning pattern. To accomplish these ends we have
resorted to the use of identical photo-images in both channels of EROS. This is
the best registration condition that one could expect to achieve in that the
images are not dissimilar in geometry as is the case with stereo imagery.
The first test was performed in the following manner. With EROS correla-
ting normally the transport was moved until one of the selected image points was
coincident with the measuring marks. At this time the EROS correlator was dis-
connected and the operator performed ten consecutive registrations at the point
involved. The precision of these settings is the best that the operator could
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
11
r
t
I
L
t
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
r
r
be expected to achieve in that the disconnection of the correlation system has
removed all optical and mechanical motion errors from the system. This operation
was repeated at each of the selected points.
One would expect that the next lower level of precision would be
achieved when the operator takes multiple readings at each point with the correla-
tor and servo loops operating. Thus, this was our second set of readings.
The third level of precision to be expected and the condition encountered
which defines the "zero parallax" readings is that wherein the operator takes
readings at each consecutive point and circulates through the points to achieve
the number of repetitions required. This procedure insures that such transport
errors as may be expected to degrade stereo measurements are brought to view and
considered in the analysis. The average parallax readings at each point result-
ing from this operation are the reference against which we gauge stereo parallax
readings. The difference between the two averages at each point is the averaging
error caused in the automatic registration by parallax anomalies within the scan
area.
2.2.4 Clearance Error Measurements
Two types of measurement data were taken by two different measurement
means in the clearance error measurements portion of the experimental program.
The primary data taken was that relating to the differences between automatic
registration of the images and the manual registration of the images. As
previously alluded to, these data were taken by circulating among the selected
points with EROS in a closed loop registration condition and at each point having
the operator adjust the x-position of the right hand measuring mark until the
floating mark in the stereo view was in contact with the point. The operator then
recorded the micrometer reading corresponding to the measuring mark position as
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
a consequence of manual registration. This circulatory mode of data recording
was continued until ten observations had been made at each point selected in the
photographs.
The second type of measurement accomplished was performed on the photo-
graphic images using a mirror stereoscope and parallax bar. These measurements
were performed to define the amount of parallax introduced by the height anomalies
within the scanned area. By taking this data we then had at hand the location
of the height anomalies within the scanned area by reference to the photographic
image and also the magnitude of those height anomalies. The intent here is to
attempt to relate the location and magnitude of height anomalies to the regis-
tration error engendered by that height anomaly at the center of scan.
2.2.5 Experimental Results
The final task in the experimental program was to reduce the data taken
into a form susceptible to analysis. For instance, this part of the program
comprised the reduction of zero parallax readings and standard errors of settings.
Additionally, the standard errors of setting were converted into standard errors
in heighting in the terrain so that their magnitudes became relatable values.
The other task that was performed in this portion of the program was to
prepare the photographic materials which illustrate conditions selected and tested.
2.3 Image Point Selections
In keeping with the time and funding restraints of the program, only
six points were selected for investigation at the two zoom settings and three
scale conditions. Figures 2.3a and 2.3b are five times enlargements of two areas
of a 1:10,000 scale photograph upon which are marked the specific points at which
observations were taken. It should be noted that at this enlargement, the scan-
ning raster sizes at minimum and maximum zoom were 130mm square and 40mm square,
respectively. (See also Appendix B.)
1
r
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 s
s
1
t
t
I
a
e
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Figure 2.3a - Location of Image Points
(5x enlarged portion of 1:10,000 scale photograph)
11
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Figure 2.3b - Location of Image Points
(5x enlarged portion of 1:10,000 scale photograph)
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
11
1
1
 I
t
t
a
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
The points selected are of three different characters. Point one was
selected atop an upward projecting building. The error here is between both the.
top of the building and the scan pattern and between the scan pattern and the
ground. Points 2, 3, 5, and 6 were chosen so that height anomalies lifted the
scan pattern from the ground while the point of interest was. on the ground.
Point 4 was chosen on a relatively planar, but sloped portion of terrain. This
selection should provide minimum error between correlator and operator settings.
Conformance of results to plans for points is discussed in Section 2.4.
2.4 Data Evaluation
2.4.1 Measuring Mark vs. Raster Relationship
As discussed in Section 2.2.1, a square bar pattern was used to evaluate
the position and variability of position of the stationary measuring mark with
respect to the center of the scan pattern area. Owing to poor video quality in
a photographic sense the polaroid test photos are not included herein, but
Figure 2.4.1 depicts the character of the photographic results.
4- Imaged Bar Pattern
Measuring Mark
Projection -
Measured Center of
Scan Raster
Figure 2.4.1 - Illustration of Measuring
Mark Projection Test Images
13
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
The measurements of these images showed that for all trials the measuring
mark did not depart from the raster center by more than 3% of the raster width.
This amount of displacement is insignificant in the tests performed.
2.4.2 Raster Size Determination
As discussed in Section 2.2.1, a double bar pattern was used to determine
scan size at the photographic image plane. Figure 2.4.2 illustrates the type of
Polaroid photographic image used in this evaluation.
Figure 2.4.2 - Illustration of Raster Size
Test Images
The known bar-space pattern in each image having orientation in two
directions allowed the evaluation of scan dimensions. It was found that at minimum
and maximum zooms respectively, the scan sizes at photo scale were 26mm and 8mm.
These sizes define the area scanned around a test point in the photo imagery
regardless of photo scale.
2.4.3 Measuring Mark Motion Calibration
Measuring mark motion calibration in the plane of the photo images was
accomplished using bar patterns oriented first in x and then in y direction. The
right (movable) measuring mark was set visually upon the edge of a bar. A
micrometer reading was taken and the measuring mark moved over one bar-pattern
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
t
I
 b
w
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
cycle to the edge of the next bar. The difference between the micrometer reading
at the second bar edge from that at the first related to the actual bar-pattern
cycle dimension provided the conversion factor from the micrometers to the photo
scale. It was found that the conversion factor between micrometer readings and
photo distances is 1.77 inches/inch at minimum zoom and 0.55 inches/inch at maxi-
mum zoom.
2.4.4 Parallax Clearance Precision
As discussed in Sections 2.2.3 and 2.2.4 and depicted in Figure 2.2,
I
I
i
identical and stereo images were registered under low and high zoom conditions,
three system conditions and three photographic scale conditions. In all 2160
parallax clearance observations were accomplished in compiling the statistics
needed in the program.
Tables 2.4.1a, b, and c show the results of precision (repeatibility)
derived from multiple observations. It will be noted that each table relates
to one photographic scale. Within the tables are recorded the standard errors
of observations (in microns) for the matrix of conditions established.
With respect to the tabulated data, certain observations appear in
order. They are as follows:
- As expected y-parallax clearance is less accurate than x-parallax
clearance, since dove prisms are not available in EROS to switch
y-parallaxes to the x-direction.
- Precision decreases when changing from the most quiescent mode (open
loop) of operation to the more dynamic mode which combines closed loop
operation with transport errors.
- Precision increases under the effects of higher system magnification.
- Precision of registration is a function of photo-machine resolution
and not a function of photographic scale.
15
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Table 2.4.1a - Standard Errors of Observations
(At photographic scale in microns)
Closed Loop
Scale 1:10,000
Open Loop
Closed Loop
and Transport
a
4
7.5
7
x
Low Zoom
aY
9
7
7
Identical
Plates
ax
2.5
4
5.5
High Zoom
a
3
4
9
Y
ax
6
6
8
Low Zoom
aY
7
9
11
Stereo
Plates
a
2
5
6
x
High Zoom
a
4.5
5
6
Y
- - " - W" W - - MOM, s w *0 "
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010039-8
Table 2.4.1b - Standard Errors of Observations
(At photographic scale in microns)
Closed Loop
Scale 1:20,000
Open Loop
Closed Loop
and Transport
ax
6
7
8
Low Zoom
a
7.5
9
11
y
Identical
Plates
OX
3
5
4
High Zoom
a
2
5.5
6
y
ax
5
7
6
Low Zoom
a
6
8
9
y
Stereo
Plates
aX
2
4
5
High Zoom
ay
3.5
4
6
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Table 2.4.1c - Standard Errors of Observations
(At photographic scale in microns)
Closed Loop
Scale 1:30,000
Open Loop
Closed Loop
and Transport
aX
5
5
9
Low Zoom
5
7
11.5
a
Y
Identical
Plates
ax
1.5
3
3
High Zoom
a
3
3
4
Y
a
4
4
8
Low Zoom
X
a
6
6
11
Y
Ste
eo
r
Plates
aX
2
4
6
High Zoom
a
3
3.5
8
Y
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
I
t
a
I
- Precision of registration of the human operator is not significantly
dependent upon whether the images are identical or stereo.
- The average x-parallax clearance precision of the operator using EROS
(open loop) with stereo photographs is five microns and two microns
respectively, for low and high zoom conditions.
- The average x-parallax clearance precision of the operator-machine
system with stereo photographs is seven microns and six microns respec-
tively for low and high zoom conditions.
- The average precision changes indicate that vibration, correlator and
transport errors add about five microns to the standard error of regis-
tration in EROS. Since this addition is nearly the same at both
zoom conditions, it is evident that accidental transport errors are
larger than correlator errors. Otherwise, there would be a similar
difference in precision at the two zooms as for the human operator.
This suggests that the right plate dc-servo motor system does not operate
with sufficient precision to allow better registration mensuration results.
However, since EROS was not designed as a mensuration instrument this
lack of precision in the servo loop is not unexpected.
2.4.5 Stereo Parallax Errors
Considering the foregoing, it may be stated that measurements of stereo
parallaxes have a precision in the order of seven microns at the input photo scale.
This means that at the photo scales used, the elevation determination precision
on the ground was:
(1:10,000)
(1:20,000)
(1:30,000)
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
These values are computed by the well known approximate scaling equation
oh = b' oP
wherein H is the flying altitude, b' is the stereo base distance at photo scale,
op is differential x-parallax, and oh is the corresponding differential height
on the ground.
t
1
With respect to the difference in parallax values between registrations
of identical plates and stereo plates, Table 2.4.2 depicts registration differences
in feet-of-height at the ground. These heights were computed by the same approxi-
mate equation as described above. It should be noted that the difference in
height is that difference between the scan pattern's virtual location with respect
to the terrain and the measuring mark as set on the point by the operator.
Referring to Table 2.4.2 and Figures 2.3a and 2.3b, the following obser-
vations are made with respect to parallax differences:
- Point 1 is atop one of a group of buildings. However, the scanning
pattern at all acales and zooms is much below the top of the building.
Especially at lower zoom and smaller scales, the image information on
the ground is much stronger than that at the tops of the buildings in
the area, thus the scanning pattern closely approaches being on the
ground. At high zoom where less area is covered by the scan raster, it
is noted that the raster climbs up the building considerably.
- Point 2 is on the ground among trees. Table 2.4.2 figures show that
the trees and buildings surrounding the point raise the raster from the
ground. Only at smaller scales and low zoom does ground information
bring the raster to the ground.
- Point 3 is on the street in a residential area. At this point it is
I
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Table 2.4.2 - Area Correlation Height Errors
(Feet on the ground)
Photo Scale
Point
1:10,000
1:20,000
1:30,000
Low Zoom
1
- 16
- 21
- 10
2
+ 6
+ 2
.+ 1
3
+ 7
+ 6
+ 10
4
+ 1
+ 6
+ 13
5
+ 1
+ 1
0
6
+ 1
+ 5
+ 6
Ave/ h
0/8
0/9
0/14
High Zoom
1
- 13
- 12
- 11
2
+ 6
+ 8
+ 9
3
+ 8
+ 9
+ 8
4
- l
- 5
- 2
5
+ 2
0
- 4
6
- 2
0
0
Ave/ h
0/7
0/7
0/7
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
seen that the averaging effect of the correlator is quite different
from that at point 2. At this point the raster is considerably off
the ground at all scales and zooms.
- Point 4 is in a sloped open field. The errors at this point occur
as would be expected. At higher zoom and larger scale, the errors
become quite small. Only when large areas are covered by the raster
does error increase.
- Point 5 is at the edge of a road where hard x-image information is
available. As a consequence it will be noted that errors are small
under all conditions of averaging.
- Point 6 is in a sand and brush area of high contrast. Here again
local image conditions appear to keep error low especially under high
zoom conditions where peripheral conditions are excluded.
It is interesting to note in Table 2.4.2 that the average error at the
six points under each condition is approximately zero. With respect to the
standard error at the six points, the average
the smaller the area on the ground represented by the scan raster the smaller
will be the error. This confirms the philosophy used in automated stereo plot-
ting instruments wherein scanning rasters are made quite small to avoid averaging
errors when only first order transformation capability is present in the instru-
ment. The disadvantage of the area reduction approach is that at some point
reliability of correlation becomes impaired.
3.0 CONCLUSIONS AND RECOMMENDATIONS
The above described test program has been sufficient to provide the basis
for four conclusions concerning the EROS instrument. These conclusions are as.
follows:
22
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
e
1
1
I
t
i
f
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
t
t
t
- The optical system has sufficient resolution to support manual
measurements in the micron range.
- The registration precision of the correlator is something better than
?7 microns at photo scale.
- The introduction of measuring marks at the eyepiece is a workable
method for a mensuration instrument if the photo stages are supplied with
micron accuracy mensuration capability. The summation of plate and
micrometer measurements would provide accurate results regardless of
the inaccuracies of the transforming optics included between the photo
and the eyepiece.
- The correlator itself cannot be used as a means to derive measurements
of height of buildings, etc., when only first order transformation capa-
bility is present in the system.
On the basis of the results of this study it is recommended that no
further testing of this specific nature be accomplished. The capabilities and
limitations of instrumentation similar to EROS have been established in sufficient
detail for future system engineering purposes.
23
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
s
1
a
1
1
t
t
t
I
Appendix A
Electronic Registration Optical Stereoscope
(EROS)
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
t
t
Page
1. System Description . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Optical System . . . . . . . . . . . . . . . . . . . . . 1
1.2 Scanning System . . . . . . . . . . . . . . . . . . . . . 3
1.3 Correlation System . . . . . . . . . . . . . . . . . . . 4
1.4 Servo System . . . . . . . . . . . . . . . . . . . . . . 5
2. Performance of EROS Feasibility Model . . . . . . . . . . . . 9
2.1 Optical System . . . . . . . . . . . . . . . . . . . . . 9
2.2 Scanning System . . . . . . . . . . . . . . . . . . . . . 10
2.3 Correlation System . . . . . . . . . . . . . . . . . . . 10
2.4 Viewing Tests . . . . . . . . . . . . . . . . . . . . . . 13
3. Requirements for Future Equipment . . . . . . . . . . . . . . 21
3.1 Optical System . . . . . . . . . . . . . . . . . . . . . 21
3.2 Scanning System . . . . . . . . . . . . . . . . . . . . . 22
3.3 Correlation System . . . . . . . . . . . . . . . . . . . 23
3.4 Servo System . . . . . . . . . . . . . . . . . . . . . . 24
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 26
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010039-8
t
1
1
t
1
1
t
1. System Description
The EROS equipment was built as a feasibility demonstration model to
prove the feasibility of combining electronic automatic registration techni-
ques with a purely optical viewing system. The stereoscope was built as a
breadboard with no intention of its being used as an operational instrument.
Off-the-shelf components were used as far as possible, and the equipment
was designed to accept input transparencies on 9-1/2 x 9-1/2 glass plates,
using an air bearing transport system.
1.1 Optical System
The optical system is based on four major components which between
them introduce the four first-order geometrical distortions: (1) Zoom
Lens; (2) Rotation K-mirror; (3) and (4) Anamorphic lenses. The four
first-order distortions can be defined in various ways. The simplest
definition is in terms of X and Y components i.e.,
X scale
Y scale
X skew
Y skew
This method is inconvenient when dealing with an optical system, be-
cause most optical components operate on a two-dimensional image field.
The distortion parameters used in EROS are therefore
Uniform Magnification
Rotation
Anamorphic Magnification Magnitude
Anamorphic Magnification Direction
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Uniform magnification is implemented by two Angenieux 100-400mm f/5.6
Zoom lenses, one in each leg. These lenses were originally designed
for use with TV image orthicons and are computed for infinite conjugates.
In the EROS equipment it is consequently necessary to combine the zoom
lenses with collimating lenses, for which purpose the 480mm f/4.5
Schneider Xenar was selected.
Image rotation is accomplished by means of two K-mirrors, one in
each leg. Anamorphic magnification is provided by four anamorphic
adapters, two in each leg. These lenses designed for use as Cinema-
scope adapters in conjunction with a regular projection lens, and mag-
nify the image by 2x in the horizontal direction and by lx in the ver-
tical direction. Each anamorphic consists of a positive and a negative
cylindrical lens (achromatic doublets or triplets), the spacing between
which is extremely critical. These anamorphic adapters normally operate
with nearly collimated lights and to use them in the present equipment
a collimator and decollimator must be employed. Optical bench tests
at the start of the program determined that achromats of about f/7
aperture were adequate for use as collimators, as they possessed signi-
ficantly lower spherical aberration over the small field angle required
than the 480mm Xenar camera lenses. Also, it was determined that the
best resolution was obtained with the anamorphic lenses back-to-back.
The complete optical system is shown in Figure 1. The light sources
consist of 30 watt 115-volt projection lamps with translucent diffusing
screens mounted on the condenser lenses. The lamp brilliance is con-
tinuously adjustable by means of an SCR control circuit.
I
1
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Alignment and scale
Input material corrected
Skew distortion
removed
0
Video
amplifier
e I I Input film
%-Y position Servo
Photomultiplier
Analog
computer
Servo
I
Waveform
generator
Servo
Dichroic mirror
Anamorphic lenses
EROS feasibility demonstration model, functional diagram
FIGURE 1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Before entering the zoom lens, the input image is magnified by
2.5x by means of a Schneider Comparon 100mm f/4 enlarging lens,
specially computed for this magnification. Simple field lenses are
provided at the two image planes in the system.
The eyepieces used are lOx B & L wide field. Interocular adjust-
ment without change of focus is provided by relative movement of both
the eyepieces and the 450 diagonal mirrors. Provision is made for
removing the eyepiece assembly and replacing it with a Polaroid back
in order to make a hard copy of the viewed images.
1.2 Scanning System
Scanning the two input transparencies is achieved through an
optical duplexing system. The scanning pattern originates on the
face of a 5ZP16 CRT and is inserted into the optical system by means
of pellicle beam splitters with dichroic coatings. The P16 phosphor
peaks at around 4,000A; the dichroic mirror reflects 90% of the light
below 4,500A and transmits 85% of the light above 5,000A. The scanning
light passes through the optical system in the reverse direction to
the viewing light causing the scanning spot to be reimaged on the
transparency. The blue scanning light transmitted by the transparency
is separated by a further dichroic mirror, collected by a condenser
lens and is made incident on the photocathode of a photomultiplier
tube which converts the light fluctuations into a video signal. To
prevent scattered light from the projection lamp from entering the
photocell, the projection light is filtered with a Wratten No.12
minus blue filter and a further No. 47B blue filter is placed in
front of the photocell.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
a
1
1
t
1
t
t
1.3 Correlation System
The correlation system employed in EROS is based on that developed
for the ARES viewer. A Lissajous crossed diagonal scanning pattern of
64 lines is employed, this giving adequate resolution for image correla-
tion. The line frequency is approximately 3 kHz using 30 frames per
second with an interlace of 2:1. The scanning waveforms are derived
from a 780 kHz 1-c oscillator. The output of this oscillator is first
divided by 2 and then divided by ratios of 63 and 64.
A further division by 2 reduces the square wave frequency to about
3 kHz. The triangular scanning waveforms are generated by an integrating
circuit. The deflection amplifiers have current feedback by which means
the output current waveform applied to the CRT deflection yoke is forced
to follow the amplitude of the input voltage.
The resistance of the deflection coils was compensated by introduc-
ing a step into the scanning waveform, so that a good triangular current
wave was obtained.
. The correlators and distortions analyzers followed the 1965 ARES
design except that all frequencies were reduced to 1/5 because of the
reduction in line frequency from 15 kHz to 3 kHZ. Only two frequency
bands were employed, 21.5 - 46.5 kHz and 46.5 - 100 kHZ respectively.
The distortion analyzer gave the normal zero - and first order
error signals:
X parallax
Y parallax
X scale
Y scale
X skew
X skew
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1.4 Servo System
1.4.1 X and Y Parallax
All image distortions are compensated mechanically or optically
in this instrument. X and Y parallax are corrected by means of two DC
Globe motors mounted on the air bearing film transport. The servo
amplifiers are operated directly from the X and Y parallaz output of the
distortion analyzer. The R-C filter time constant at the output of the
distortion analyzer must be adjusted to eliminate most of the correla-
tion noise, as this can easily saturate the servo amplifiers, resulting
in decreased loop gain.
1.4.2 Zoom Lenses
The uniform magnification error signal is obtained by adding the
X scale and Y scale errors. It is then applied in opposite polarity to
the servo amplifiers driving the two zoom lenses, so that the difference
in magnification is taken out equally on both sides. In addition, to
responding to differences in magnification between left and right images,
the zoom lenses must also accommodate the manual zoom control. This is
achieved by using a position servo with follow-up pots which controls
the average position of the two zoom lenses. The manual zoom and auto-
matic differential magnification controls remain completely independent
because the average position is unaffected by equal positive and nega-
tive changes in the two lenses.
In the manual mode, the differential magnification is controlled
from a front panel potentiometer with position feedback derived from
the difference between the outputs of two follow-up pots on the zoom
lenses.
t
t
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 1
t
a
t
t
t
t
t
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1.4.3 K-Mirrors
The rotation error is given by the difference between the X-skew
and Y-skew errors, and is applied to the K-mirror servo amplifier. The
servo motor, a Printed Motors Incorporated Model, drives the two K-mirrors
in opposite directions through a gear ratio of 89:1
In the manual mode, image rotation is controlled by a front panel
potentiometer using a follow up pot connected to the K-mirrors.
1.4.4 Anamorphic Lenses
The two remaining first order distortions are corrected by the
anamorphic lenses. The anamorphic distortion can be specified in two
ways: (1) in polar coordinates as the magnitude and angle of the
anamorphic stretch; (2) in cartesian coordinates as the magnitudes of the
differential scale and skew components. The latter can be obtained
directly from the distortion analyzer output:
differential scale error A b = X scale error - Y scale error.
skew error Ac = X skew error + Y skew error.
The two anamorphic lenses do not operate independently in correcting
their two distortion components. The required position (or motion)
of each anamorph depends on four factors:
(1) Its own orientation
(2) The orientation of the second anamorph
(3) The differential scale error
(4) The skew error
Thus it is necessary to solve two simultaneous equations in order
to determine in which direction each anamorph should be rotated to
eliminate a given distortion. If an anamorph is rotated in the wrong
A-11
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
direction, then the feedback in the correlation loop becomes positive,
resulting in oscillation or driving the anamorphs to their extreme
position.
The equations to be solved are
. =Ab cos 22 +Ac sin 26
a 1.5 sin 2a-R
Abcos2a+Acsin2
1.5 sin 2 a -s
where Ab = differential scale error (x scale - y scale)
Ac = skew error (x skew + y skew)
a = angle of first anamorph
13 = angle of second anamorph
As the denominator goes to zero when a =a these equations are
difficult to instrument. In EROS, the simplification of using only the
sign of the trigonometric functions was used, resulting in the equations:
a = SIGN (a-s) [ Ab coSIGN 2s + Ac SIGN 2s
SIGN (a-s) ' A b coSIGN 2 a +A c SIGN 2a
where SIGN 2a is defined as sin 2a
Isis 2al
coSIGN 2a is defined as cos 2a
Icos 2a l
The analog computer used to solve these equations uses sine/
cosine potentiometers coupled to the anamorphs to derive the trigono-
metrical functions, the outputs of the potentiometers being amplified
and hard limited to preserve only the polarity of the signal. These
signals operate relays which switch when the angles change sign.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
1
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
e
1
t
f
1
t
a
t
The effectiveness of the anamorphic correction varies a great
deal with different types of material. The gain required in the anamorph
loops varies with the information content of the material viewed. Thus,
a gain setting that is satisfactory in one area may cause oscillation
in another; if the gain is reduced to avoid oscillation, then there may
be insufficient correction in the first area. In practice, a compromise
gain setting can be made that will operate with most input material.
A-13
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
2. Performance of EROS Feasibility Model
2.1 Optical System
2.1.1 Magnification
The magnification of the optical system is variable over a range
of 4:1. Using lOx eyepieces, the effective magnification is 6x to 24x.
2.1.2 Field of View
Field of view with the lOx eyepieces varies from 32mm at 6x to
8mm at 24x with the anamorphs set in the neutral position. When the
anamorphs are set to give approximately 2:1 anamorphic ratio, the field
of view at 6x magnification is approximately 2.2 x 4.4mm. When the
anamorphs are set to give 4:1 anamorphic ratio, the field of view is
1.6 x 4.4mm, (Limited by the diameter of the first field lens).
2.1.3 Exit Pupil
The exit pupil is normally 3mm dia. with a relief of 18mm at
minimum magnification and 2.5mm dia. with the relief of 18mm at maximum
magnification. At 4:1 anamorphic ratio, the exit pupil is elliptical
and measures 4 x 2mm.
2.1.4 Illumination
The illumination level in the film aperture is variable up to
about 240 foot candles (open gate). This gives an illumination level
of the image in the eyepiece of 1.2 ft-c at minimum magnification and
.36 ft-c at maximum magnification. Comfortable viewing with a lOx eye-
piece in ambient room light requires about 0.5 ft. candle illumination
in the final image.
1
t
t
t
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
e
t
t
I
t
2.1.5 Resolution
The limiting resolution as measured by observation of a USAF
high contrast resolution target at maximum magnification using the lOx
eyepieces is 144 lines/mm.
2.2 Scanning System
2.2.1 Cathode Ray Tube
The Scanning CRT was operated at 20kv, with a cathode current of
50~uA. The raster size was 0.8 x 0.8 inch.
2.2.2 Video
Video and noise signal levels at the output of the video amplifier
were as follows:
Left Side Right Side
Noise Voltage (Average) 0.lv 0.2v
Peak Signal Level (Clear Aperture) 1.2v 2.Ov
(Typical Photo) 0.3 - 0.5v 0.3 - 0.5v
2.3 Correlation System
Using two correlators covering the Video frequency band of 21.5 -
100 kHZ, performance of the correlation system was measured with a stereo
pair of high resolution panoramic inputs on black and white film. The
photographs were first automatically registered in the normal manner and then
all the servos were disconnected. The open loop transfer characteristics
were then measured one at a time by moving the appropriate optical component
from its set position and recording the error voltage at the servo amplifier
input. Between each measurement, the system was re-registered automatically.
A-15
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
2.3.1 X and Y Parallax
Minimum magnification - Field of view 32mm.
Error Voltage - ? 1.0 volt for ? 1.5mm parallax
Error Constant - 0.7 volts per mm or 0.225 volts per cent displacement
Pull in Range - Approximately ? 2.5mm
Noise Level - Approximately .05 volt rms.
2.3.2 Zoom Lenses
Magnification approximately 10x.
Error Voltage - ? 5.0 volts for ? 5 per cent magnification change.
Error Constant - 1.0 volts per cent
Pull in Range - Approximately ? 15 per cent
Noise Level - Approximately 0.1 volt rms.
2.3.3 K-Mirrors
Error Voltage - ? 2.5 volts for ? 1/4 degree image rotation.
Error Constant - ? 10 volts per degree image rotation
Noise Level - Approximately 0.1 volt rms.
Note: K-mirror rotation angle is 1/2 image rotation.
2.3.4 Anamorphic Lenses
Two error signals are corrected by the anamorphic lenses, differential
scale ob, and skew, cc. The four anamorphic lenses are cross coupled in
pairs on left and right viewing channels, resulting in only two anamorph
control signals.
A change in the orientation of either pair of anamorphs produces both
ob and [c errors.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
s
t
t
1
t
t
t
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
1
t
t
The error voltages obtained when the anamorphs are displaced depend
greatly on the quality and geometry of the imagery. The following results
were obtained with convergent panoramic material of low contrast but with
good imagery. The angular separation of the anamorphs was approximately 5?
and the cylinder axes were aligned at about -10 S' to the horizontal axis,
i.e., the distortion being corrected was mainly skew.
Anamorph Angle L b c
Degrees Volts Volts Volts Volts
-2.4 -?20 +?70 -?90 -?65
0 -.15 -?05 0 -.10
+2.4 +?20 -.75 +1.15 +?80
0 0 0
+2.4 -1.0 -?75
Noise level (average value measured with DC Meter) 0.lv peak to peak.
From these figures, the correlation error constant fora ands
can be estimated at about 0.4 volts per degree of anamorphic lens rotation.
This sensitivity was much higher than required; in practice the loop gain
was reduced to about 1/5 in order to obtain stable operation.
A-17
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
2.4 Viewing Tests
To determine the overall correlation and scanning capabilities of the
EROS viewer, tests were conducted with a variety of input materials.
The types of imagery used for these tests included
1. 92 x 92 inch wide angle convergent frame.
2. 70mm high altitude oblique panoramic (black and white)
3. 70mm high altitude oblique panoramic (color)
4. High altitude vertical panoramic (KA 58)
5. Low altitude vertical panoramic (KA 56)
6. Low altitude oblique frame (KA 51A)
7. High altitude oblique frame
2.4.1 9,12 x 92 wide angle convergent frame photography.
The convergence angle on this material is ?306, and the scale
difference between corresponding images at the edges of the left and right
frames is about 2.5:1. It is therefore a very difficult test of image
correlation and was the primary test material used in the development of
the equipment.
The EROS viewer will correlate and track this material from edge
to edge, providing a good stereo model. Because of the equal convergence
angles, a near vertical apparent viewpoint is obtained over most of the
frame. Some difficulty is experienced in correlating areas of almost
featureless desert and scrub due to the lack of significant imagery.
Mountainous areas also are difficult because the large convergence angle
(60' between viewpoints) presents totally different image detail in the
left and right frames. Because of the rapidly changing distortion (the
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
i
t
t
t
 1
1
t
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
anamorphic lenses rotate about 900 from edge to edge) the imagery must
be scanned quite slowly. This is due mainly to the narrow bandwidth of
the servo system, necessitated by the low signal-to-noise ratio of the
error signals delivered by the correlator and analyzer. With the improved
designs being developed, the scanning rate of future equipment should be
considerably increased.
While this imagery is of low resolution (20 - 30 lines/mm) and
cannot do justice to the resolution capability of the EROS optical system,
it does demonstrate the ability of the equipment to correct automatically
considerable geometrical distortion.
2.4.2 High Altitude Oblique Panoramic (black & white)
This material was obtained with a 13" lens at a tilt angle of
?72 degrees. The imagery consisted of a large city and dock area, partly
covered by thin cloud. Good correlation was obtained in all areas except
where more than about 50% of the frame was covered by water, and the strip
could be scanned quite easily, even into the partially obscured area.
2.4.3 High altitude oblique panoramic (color)
This material was obtained with the same taking parameters as
The imagery contained good detail consisting of a city and some
wood e d hills. Good correlation was obtained and the film could
be scanned quite rapidly. Because the optical duplexing system removes
blue light from the viewed image, the image had a yellow tint but was
very pleasing to look at.
A-19
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
2.4.4 High Altitude Vertical Panoramic
This material was obtained from the KA 58 camera which is of 18
inch focal length with a scan of 1400. This material correlated and
scanned easily. Alternate panoramic strips, covering the same horizon,
were used as stereo pairs, and good results were obtained to within 2
inches of the horizon. At this point the image contrast on film is very
low, and the limitation appears to be due to lack of image contrast rather
than any lack of distortion correction capability.
2.4.5 Low Altitude Panoramic
This material came from the KA 56 camera of 3 inch focal length
with a scan of 180'. Altitude was 2900 feet. Good results were obtained
with this material, but the geometrical distortion changes very rapidly
in this format, and it was necessary to scan the material fairly slowly.
Correlation could be obtained in most areas, but the best scanning was
obtained at middle distances (around 45` oblique). It is believed that
this is due largely to the imagery rather than the geometry of the format.
Near nadir the scale is large and there is little detail on which to
correlate (a field of corn covers the whole viewed area, for example,
even at lowest magnification.) Near the horizon the image contrast is
low.
2.4.6 Low Altitude Oblique Frame (Side looking)
This material was obtained with the KA 51A camera of 6 inch focal
length. Results with this material were surprisingly good. The photo-
graphs contain high order distortions that cannot be corrected by EROS,
but nevertheless an interesting stereo model was obtained.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
1
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
Major objects such as houses were transformed to their correct
average position, but were individually somewhat skewed. This resulted
in a stereo model in which nearby vertical objects leaned over. Distant
objects appeared normal. The entire photograph could be scanned quite
rapidly and stereo accommodation was obtained in all areas.
2.4.7 High Altitude Oblique Frame
This material was obtained with a 36 inch F. L. lens at various
angles of obliquity. Chips from both vertical and oblique frames cor-
related and scanned easily.
A-21
STAT
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
1
2.4.9 General Observations
The above tests have shown that the EROS viewer will operate
satisfactorily with a wide variety of input materials. Performance was
better with some types of photography (notably panoramic) than with
others, but in no case was there failure to obtain correlation due to
image geometry. With non-photographic inputs
able difficulty was experienced due to the presence of higher order dis-
tortions. The anamorphic stretch capability of 4:1 in each channel was
more than sufficient to cope with any distortion encountered. In fact
a ratio of 2:1 in each channel would probably be adequate. It would be
desirable to increase the zoom range to 10:1, to provide more manual zoom
range after accommodation of the difference between the scales of the two
inputs. The range should.be extended at both ends so that fields of view
both greater and smaller than at present can be accommodated. This would
enable the image content of any area to be adjusted for optimum correlation
t
STAT
t
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8 I
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
I
t
1
t
over a wider range of photographic scale factors. This is especially
desirable for low altitude panoramic where the scale changes considerably
from nadir to horizon.
As with the ARES viewer, there are a few areas of imagery in which
correlation is difficult to obtain. The main causes of poor correlation
and their solutions appear to be as follows:
(a) Lack of image detail within correlator passband.
(b) Low contrast imagery.
(c) Different appearance of imagery in left and right channels due to
viewpoint, specular reflections, shadows, etc.
(d) Large difference in image detail in X and Y directions.
Lack of image detail within correlator passband.
This can be remedied by increasing the zoom range, increasing the
correlator bandwidth, or both. The use of a tunable correlator which
selects the frequency band containing most information would be an ob-
vious improvement here. However, it must be borne in mind that the low
frequency end of the video spectrum is of greatest importance in estab-
lishing initial lock-on.
Low contrast imagery.
The imagery may be of low contrast at any density level. Hence it is
necessary to have a video sampling system that will cope with large varia-
tions in light level and contrast. This can be achieved by an automatic
gain control amplifier that maintains a constant amplitude AC output and
suppresses the DC component.
A-23
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
Different appearance of imagery.
When the imagery in left and right channels is appreciably different,
the output of the correlator degenerates into noise, which tends to drive
the distortion correction components in random directions. The only solu-
tion therefore is to monitor the correlator output, and to inhibit operation
of the automatic controls when the signal to noise ratio drops below a
predetermined figure. This will hold the distortion corrections at their
previous value through areas of poor or different imagery. Automatic
operation can be resumed without interruption providing that the distortions
are still within the lock in range of the system when good correlation
is re-established.
Large difference in image detail in X and Y directions
The amplitude of the error signals delivered by the distortion analyzer
depend not only on the actual geometric image distortion but also upon the
amount of image detail present. One-dimensional image structure is fre-
quently present in aerial photography, e.g. highways, railroads, canals,
etc., which produce much higher error signals in one direction than another,
depending on their orientation. This creates a special problem with the
present correlator and analyzer because two linear distortion errors must
be combined to determine each of the two-dimensional optical corrections.
For example, the optical rotation error is found by taking the difference
between the X skew and the Y skew errors. If there is appreciable dif-
frence between image detail in X and Y then the X skew and Y skew error
voltages may be different even if the geometric distortion is the same in
each direction. The result is that in addition to the wanted skew distor-
tion error, a spurious rotation error is indicated.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
1
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
t
t
1
An unbalance in image detail can similarly produce spurious indica-
tions of magnification or anamorphic error. In attempting to correct
such non-existent errors, the system may drive out of lock.
The solution to this problem lies in using a distortion analyzer
which is better matched to the optical distortion correction parameters.
This involves using the correlator output only at those instants in which
the scanning spot is moving in a direction in which the optical component
can correct image displacement. For example only radial parallax should
be used in determining the magnification error, and only rotational
parallax should be used in determining the rotation error. A new distortion
analyzer based on these ideas is now being developed and will be used on
future equipment. A considerable increase in the reliability of correla-
tion should result from this development.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
3. REQUIREMENTS FOR FUTURE EQUIPMENT
Development of the EROS breadboard feasibility model has indicated many problems
that must be solved in future equipment. These will be briefly described
under the headings of the four main subsystems.
3.1 Optical System
3.1.1 The moving elements of the zoom lenses were supplied with con-
siderable clearances (probably to ensure that they never seized up under
any circumstances) with the result that random image shifts occur when
the lenses are zoomed. These shifts occur so rapidly that the X-Y
parallax cannot correct them; they are most noticeable at maximum mag-
nification.
3.1.2 Optical alignment of the system proved extremely difficult,
especially the K-mirror. The main problem was adjustment of the three
mirrors to obtain rotation of the image about a fixed point. In addition
to the K-mirror components there are three fixed mirrors in this part
of the optical system each with three adjusting screws which makes a
total of 18 adjustments on each side, all affecting image rotation.
3.1.3 It is not sufficient to adjust the spacing of the anamorphic
lens elements to perfect collimation (zero power) as they invariably
possess small power in the direction of the cylinder axis, due to manu-
facturing errors. To avoid astigmatism in the final image it is neces-
sary to adjust the lens spacing so that the power in the two directions
is equal. This is best done on an optical bench using a collimator
and decollimator lens on each side of the anamorph and adjusting the ana-
morph for minimum astigmatism.
t
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 r
1
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
3.1.4 Another problem encountered with the anamorphs was that a signi-
ficant shift of the image center occurs with anamorph rotation. This
is probably due to misalignment of the lens axis with the rest of the
optical system. Provision should be made in future for precise adjust-
ment of the anamorph axis.
3.1.5 Due to the presence of two field lenses, some field curvature
is apparent in the viewed image. This can be removed by a properly
designed negative element in the eyepiece itself.
3.1.6 The viewed image in EROS is inverted due to the use of an even
number of image reflections while viewing the input material from
below.
3.2 Scanning System
a
3.2.1 The P16 phosphor peaks at around 4,000A, a wavelength at which
the flint glass elements of the optical system have considerable absorp-
tion. This results in a large light loss and necessitates a slow scan
and narrow video bandwidth in order to achieve a reasonable signal to
noise ratio.
3.2.2 Due to light leakage into the photomultiplier, the brightness of
the viewing light source is severely restricted. Even with-the use of
efficient color filters on the light source and photomultiplier, this
is still a problem. The cause is not direct leakage in the lamphouse/
photomultiplier assembly (which has been virtually eliminated) but
seems mainly due to scattering of the light from the film and other
glass surfaces, many of which are plane, back into the photomultiplier.
3.2.3 Because of the yellow color of the viewing light, the viewing
of color film is impaired.
A-27
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
A-28
These problems can all be solved by the use of a scanning system
employing an image dissector or a pair of vidicons.
Correlation System
3.3.1 Correlation Noise
The major limitation to tracking speed in EROS is the bandwidth
of the correlation loop. As the automatic registration system is in
effect six phase locked loops in parallel it is necessary for the X- and
Y- parallax at least to track the error signals rapidly enough to remain
on the linear part of the loop transfer function. Preferably, all of
the loops should have this capability. Thus, the bandwidth of the corre-
lation loop is a vital parameter in system performance, and we would like
to have this bandwidth as large as possible. The main factor determining
the bandwidth of the correlation loop is the signal-to-noise ratio of the
error signals at the output of the distortion analyzer.
This is dependent on many factors such as the method of distortion
analysis, the CRT or vidicon frame rate and the video signal-to-noise
ratio.
In EROS considerable noise is present at the distortion analyzer
output and the bandwidth has to be limited to about five cycles per
second to provide a useable signal-to-noise ratio. It is most important
in future equipment to reduce the correlation noise. While a higher
video signal-to-noise ratio would be an advantage, it is believed that a
considerable amount of the noise is generated in the distortion analyzer
itself. One reason for this may be the fact that the basic scanning
motion in the Lissajous pattern is at 45`, while the distortion compo-
nents are resolved in the X and Y directions. To find X- parallax, the
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
I
1
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
1
45? scans are combined so that the X components add and the Y components
subtract. However, this results in the noise from both X and Y compo-
nents being added so that the overall signal-to-noise ratio is halved.
A further source of noise may be due to uncompensated second-and higher-
order distortion.
A further improvement would result from the use of a higher frame
rate (there being a noticeable noise component at the frame frequency),
enabling effective filtering to be done without reducing the correlator
bandwidth to a low value.
3.4 Servo System
3.4.1 The servo amplifiers used in EROS were of the. silicon controlled
rectifier type using a 60 c/s power supply. These amplifiers do not
give a smooth output due to the presence of 60 c/s pulses, and they are
tricky to operate because the control pulses cause interference to
other units. Linear DC amplifiers are recommended for future designs.
3.4.2 Printed motors were used as power transducers. While these are
potentially capable of direct drive operation at very low speeds, this
proved unsatisfactory due to the rough output of the SCR amplifiers.
Because of their very low resistance, special amplifiers are required
for printed motors. Because of this it may be preferable to use Inland
Torque Motors which can be obtained in higher resistances.
In any case, direct drive motors with tachometer feedback seem
to offer the best method of driving the optical components as this system
ensures small dead band, linearity, and absence of friction and backlash.
A-29
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
3.4.3 An effective method of eliminating the drive from a zoom lens
when it reaches the limit of its travel and transferring it to the
other zoom lens should be provided. In EROS, slip clutches are provided
which safeguard the lenses, but this results in the reductl^n of loop
gain by a factor of 2 when either lens is drive to the limit of its travel.
3.4.4 The loop gain of the X and Y parallax loops varies with the
magnification of the optical system. It is necessary, therefore, to vary
the X-Y parallax loop gain with zoom lens setting. This could be achieved
by ganging two gain pots (connected in the X and Y parallax servo loops)
with the main zoom control.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
t
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
1
4. Conclusions
The feasibility of combining an optical viewing and distortion correc-
tion system with an electronic image correlation system has been established.
Using off-the-shelf components, the resolution of the feasibility model
was 144.lines/mm at 24x magnification, and it proved capable of correcting
X and Y image parallaxes and all first order distortions to a degree adequate
for comfortable stereo viewing. The first order distortion correction
capability of the instrument was more than adequate to deal with any material
tested.
The rate at which film could be scanned varied considerably with the
type of material. High altitude panoramic material, in which the distortion
parameters change relatively slowly, could be scanned at about 1 inch per
second at 6x magnification without loss of correlation. The scanning rate
with low altitude and oblique material was somewhat lower due to the rapidly
changing distortion and the limited bandwidth of the correlation system.
The major limitations on performance as far as reliability of lock and
scanning speed are concerned lie entirely in image sampling, correlation
and distortion analysis i.e., the electronic system. Potential improvements
in this area have already been discussed in some detail. The problem of
driving the anamorphic lenses finally yielded to a simple analog computer.
No basic problems are foreseen in the optical system, and no difficulty
was experienced in driving the large optical components at the required rates
and to the necessary precision.
Development of this feasibility model has provided valuable information
on the future operational possibilities of this type of equipment. It has
provided a solid basis for future designs, and has indicated the areas of
technology where most development effort should be concentrated.
A-31
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
 STAT Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8
Next 1 Page(s) In Document Exempt
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010039-8