PANORAMIC STEREO EXPERIMENT FINAL REPORT
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP78B04747A000600140007-4
Release Decision:
RIPPUB
Original Classification:
K
Document Page Count:
17
Document Creation Date:
December 28, 2016
Document Release Date:
March 28, 2001
Sequence Number:
7
Case Number:
Publication Date:
June 15, 1962
Content Type:
REPORT
File:
Attachment | Size |
---|---|
CIA-RDP78B04747A000600140007-4.pdf | 1.3 MB |
Body:
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
Declass Review by NIMA/DOD
Approve
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
PANORAMIC STEREO
EXPERIMENT
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
INTRODUCTION Photo reconnaissance from medium and high
altitudes imposes stringent requirements on the interpretation system
if small target objects are to be recognized and functionally understood.
One of these requirements is high magnification, which implies higher
resolution in the original negative than has been available from con-
ventional cameras and films. Another is stereoscopy. As all photo
interpreters know, targets which are fairly large, simple in structure,
and familiar in appearance can be recognized by monocular viewing;
but their relative heights, details of structure and arrangement, and
(if they are unfamiliar) their probable functions cannot be deduced with
any certainty without making use of stereo viewing. The intelligence
interpreter faced with the problem of understanding an unknown structure
or other target needs every additional bit of information made available
by improved stereo techniques.
Aschenbrenner 1 and others have shown that in medium- and high-
STATINTL
altitude photography the stereoscopic effect obtained with conventional
verticals having 60 per cent overlap is at best barely adequate for the
interpretation of small, complicated, and unfamiliar objects; convergent
STATINTL
stereo pairs with a base/height ratio of three times the "normal" are
decidedly superior. In the experimental program just concluded=
have combined convergent stereoscopy with the unique
S,+TATINTL advantages of the -panoramic camera and the efficiency of projection
STATINTL stereo viewing. The has a focal length of twelve inches and can
1Aschenbrenner, Claus M. , "High altitude stereo techniques," Photo-
grammetric Eng. , Vol. 16, no. 5, December 1950, pp. 712-719.
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
thus be taken to quite high altitudes without sacrificing the resolution
obtainable with modern film emulsions. Moreover, with an appropriate
viewing device, useful stereoscopic fusion can be maintained over a
surprisingly wide portion of the panoramic scan.
The combination of panoramic configuration with converging
camera axes poses serious problems of geometry. The projection de-
vice or the interpreter's eyes, or both, must accommodate two kinds of
distortion: the increasing obliquity (and decreasing scale) toward the
edges of the panoramic strip, perpendicular to the flight line; and the
obliquity in the direction of flight produced by the forward tilt of the
camera axis (Figure 1). The experimental projection viewer used simple
mechanical means of rectifying these distortions.
Two areas were chosen for test flights: the Central Valley of
California and the west shore of San Francisco Bay. The flat surface of
the Central Valley gave a useful indication of the horizontal plane in the
projected stereo model, and the rectangular road pattern served as a
check on the rectification of tilt in both x- and y-directions. In this area
there is a variety of rural and semirural structures, orchards and field
crops, highway overpasses and bridges, whose images in the stereo
model effectively demonstrated the possibility of perceiving and measur-
ing heights. The San Francisco Bay shore includes a wide range of
natural and cultural landscapes: rolling and rugged topography, shoreline
and water features, large urban areas with many industrial and military
installations, open and dense vegetation.
The convergent coverage was obtained with a single camera,
mounted with a fifteen-degree forward tilt, by repeating each pass in the
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
STATINTL
.i
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
reverse direction. This amount of convergence provided a base-height
ratio which gave satisfactory stereo exaggeration in the projected
model: target objects could be clearly discerned and understood, and
with a suitable measuring device their heights could have been determined.
All flights were made at 20, 000 feet, which with a 12-inch focal length,
gave a contact scale of approximately 1:20, 000 at the center of format.
At 20X magnification both viewing scale (1:1, 000) and sharpness were
excellent for study of small targets.
The following paragraphs summarize the work performed by
for the Panoramic Stereo Experiment.
EQUIPMENT MODIFICATION The panoramic came r?T~,T~NTL
suitable in all respects for use in this program except that the film trans-
port and the scan rate were too slow. The film transport rate was
speeded up to five seconds. The scan rate was increased to 0. 5 second
for three flights; for the April 26 flight, on which a slow film was used,
the scan rate was slowed to about 1. 5 seconds so as to permit a longer
exposure time.
A new camera mount, combining stability with an unusual degree
of rotational freedom, was designed and fabricated. The mount provided
not only the required forward tilt of 15 degrees, but also 10 degrees of
lateral tilt correction and a full 360-degree rotation about the vertical
axis. The mount was installed in a Cessna 180 aircraft.
Since the road pattern in the Central Valley is very regular, it
was convenient to position exposures by visual observation. The flights
over the Valley were therefore planned for manual exposures. Photographs
-3-
.?+ Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
were taken one-half mile apart, with identical ground areas covered on
the return pass. An A-2 view finder was purchased and modified to fit
the space limits of the aircraft and camera mount. In addition, two pairs
of auxiliary cross-hairs were etched on the glass viewing plate, so that
at the instant of exposure the aerial photographer could identify the center
of format, not only of the photo being taken, but also the previous photo
and the photo to be taken next. This addition greatly facilitated the proc-
ess of manual exposure.
The photographs of the Bay Area were taken with intervalometer
setting, timed so that the stereo mates overlapped instead of covering
identical areas.
TEST FLIGHTS On April 13 a test flight was made over the
Central Valley. In spite of extensive ground testing of the camera before
the flight, the film transport system proved faulty and satisfactory cover-
age could not be obtained. A new cam was installed and the flight was
repeated the next day. Photographs were taken over a five-mile pass
from a point just south of Modesto. It had been planned to fly additional
photography over the Bay Area on the same day, but this could not be done
because of coudy weather near the coast.
On April 26 the third flight was made. Forty exposures were made
in each direction over an east-west strip twenty miles long in the Central
Valley. The film used was Eastman Kodak SO- 132, which has the rather
slow ASA speed of 1. 6 but the very high resolving power of 400 lines per
millimeter. Exposure time was accordingly lengthened and the scan rate
proportionally slowed. One pass was also made over the Bay shore,
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
southeast to northwest. Before the return pass could be made, however,
a second malfunction appeared in the film transport system.
.a
The camera was immediately repaired and retested, but contin-
.4 uously poor weather conditions precluded reflying the Bay Area convergent
coverage until June 12. The flight covered a strip approximately thirty-
five miles long from San Jose to San Francisco International Airport;
44 exposures were made on the first pass, southeast to northwest, and
49 on the return pass. For this flight the co-pilot's seat and view finder
were reversed so that the aerial photographer could sit comfortably,
facing aft, instead of kneeling on the floor as on the earlier flights. This
adjustment materially improved the precision of ground coverage during
long passes.
.i STEREO PANORAMIC VIEWER Two projection viewers devel-
STATINTL opted by were modified for stereoscopic projection using crossed
of
polarization. The modified viewer (Figures 2 through 6) displays the
70-millimeter panoramic negatives placed vertically in 18-inch plastic
carriers, on which guide lines are scribed for orientation. Thus the
a flight line in the projected stereo model is always horizontal. The hori-
zon in the direction being viewed is kept at the top of the model simply by
reversing the film carrier. To help the interpreter's eyes accommodate,
the top and bottom of the 45-inch square viewing screen were masked off,
making the image area 15 inches high by 45 inches wide (Figure 3).
With the modified experimental viewer, registration of conjugate
images is done manually by the projectionist. First the two photo centers
.ri
are superimposed on the screen.. Then any part of the panoramic image
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
can be brought onto the screen by means of a coupled drive mechanism
which moves the two film carriers vertically before the projection
lenses (Figure 4). Errors in registration in the direction of flight
(y-parallax) are corrected by a cam arrangement which rotates one or
both film carriers about the theoretical convergence point (Figures 5
and 6). The precision of this correction depends mainly on the corres-
pondence of ground area covered by the stereo mates; if the ground
area corresponds, then the amount of rotation required to correct
errors in y-parallax should be exactly equal for the two images. In
this experiment, in fact, only very slight differential corrections were
required. The experiment showed that satisfactory registration in the
direction of flight could be maintained throughout the image area. Al-
though perfect registration is possible only at one point, the maximum
misregistration at the edge of the 45-inch screen was only about one-
half inch, an error which the eyes aided by polaroid glasses can easily
accommodate. Even when deliberate errors of up to two inches were
made, the interpreters could fuse the images without discomfort.
Errors in registration in the direction of scan (x-parallax)
proved more troublesome. The rectification mechanism was a simple
cable arrangement which tilted the viewing screen, compensating the
obliquity produced by the panoramic scan. This method of compensation
proved adequate up to about fifteen degrees either side of the flight line.
Beyond this angle the distortion of target objects was too great for com-
fortable stereoscope fusion, and displacement of the horizontal plane
became clearly apparent (Figure 7).
,,,, Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
~S,TATINTL
On May 15 representatives of the contracting agency visited=
to inspect the stereo panoramic viewer, view sample
stereo pairs, and discuss possible uses of convergent panoramic instru-
mentation in intelligence interpretation systems.
CONCLUSION The Panoramic Stereo Experiment was not
undertaken to provide image quality consistently satisfactory for opera-
tional use,or a final viewer design; but rather to demonstrate that it is
feasible to construct a projection stereo viewer incorporating rectifica-
tion for convergent panoramic photography, and to suggest directions
that further design work may take. =believes that this purpose & TINTL
been achieved . The adaptation of existing equipment in the present
experiment was of course attended by severe limitations, of which the
most troublesome was the screen tilt resorted to for correction of
x-parallax. It is, however, entirely practicable to design a stereo
projection viewer' having more refined and reliable means of rectifica-
tion; all corrections would occur at the projector, in no way distorting
images or displacing the horizontal plane in the stereo model. With such
a projection viewer, stereo mates could be registered with a negligible
error in x-parallax (that is, no error that could not be comfortably
accommodated by the eyes) and no error in y*parallax out to 20 to 22
degrees either side of nadir. The viewer could be operated by one man,
and would permit either continuous stereo interpretation within the
1Similar in design principle to _EN-71 printer, which ?Tu#JJNTL
matically produces enlarged rectified panoramic prints.
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
angular limits just stated; or, with greater operator comfort and proba-
bly greater efficiency, monocular scanning with provision for switching
in the conjugate frame at any point where detailed stereo interpretation
is deemed necessary. The necessary controls could be built into a
panel about 8 by 10 inches in size (Figure 8).
The feasibility experiment just concluded has been based on the
most difficult geometry likely to be encountered in photo intelligence
work, and has pointed up the problems connected with further develop-
ment of the convergent panoramic idea. =belief that these progTA TL
can be solved is based not only on this experiment, but also on extensive
experience with the geometry of panoramic photography. 1 Projection
equipment which will make it possible to combine scanning and stereos-
copic study, at scales suitable for the understanding of small targets, is
needed to exploit the full information potential of panoramic reconnaissance
photography.
1See, for example, Laboratories, "Panoramic progress,"
Photogrammetric Eng. , Vol. 27, no. 5, December 1961,
pp. 747-754, and Vol. 28, no. 1, March 1962, pp. 99- 107.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
.,~ convergence lines
U11 Gl. I.1V11 ki 111 'lib
Figure 1. Distortion effects in stereo pair of panoramic photographs
with convergent tilt of 15 degrees.
Figure Z. Plan view of stereo panoramic viewer.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Figure 3. Photograph of experimental stereo panoramic viewer.
Projector heads are seen at left, masked Polacoat screen with
cable tilting mechanism at right.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Figure 4. Rear view of stereo projectors, showing coupled film drive belt.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Figure 5. Close-up of stereo projectors, showing conjugate photographs held
in film carrier and cams which rotate photographs about their theoretical con-
vergence point. See Figure 6 for geometry of y-parallax correction.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Figure 6. Rotation of film carriers about convergence point corrects
errors in y-parallax. Film carriers are viewed looking toward screen
from light source. Counter-rotation of pinion raises or lowers frame
so that convergence lines are horizontal when crossing the optical axis.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78B04747A000600140007-4
image area 45"
- flight I line - -
no error in x-parallax.
51 from 1 flight line
n
100 from Iflight line
'/
15? from X flight line
error in x-parallax can be accommo-
dated by eyes without mechanical or
optical correction
error in x-parallax can be corrected
mechanically by tilting screen
n
at angles greater than 15 degrees, optical
correction at projector head is required
for interpretability of stereo model
20? from (flight line
Figure 7. Increasing error in x-parallax with increasing
distance from flight line.
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4
Figure 8. Control Panel for Stereo Projection.
.r Approved For Release 2001/07/16 : CIA-RDP78BO4747A000600140007-4