AUTOMATIC AND SEMI-AUTOMATIC MAPPING
match electrically. To accomplish this, images
are transformed into voltage-time series.
The most exact match of the voltage-time
series determines the z coordinate or the sur-
face of the model at that xy point. There are
several ways of approaching the determina-
tion of this match. Included in these are
phase-comparison, amplitude match, and
area correlation.
At the Engineer Research and Develop-
ment Laboratories we have investigated
these three in an effort to automatically pro-
duce contours.
[n 1950, the U. S. Army Engineer Research
and Development Laboratories, Fort Belvoir,
Va., entered into a contract with Bausch and
Lomb Optical Company, Rochester, N.Y.,
to study the feasibility of producing contours
automatically. This contract followed some
preliminary experimentation by Bausch and
Lomb on a suggestion by Dr. George Harrison
of the Massachusetts Institute of Technology,
and established the possibility of automatic
contouring. With the equipment produced
under this contract, it was possible to match
successfully only artificial, high-contrast,
stereo-projected diapositives. This instru-
ment was profile follower with a phase-com-
parison match using a short multiplex bar
with two projectors. Since we did not want a
60 cycle A.C. component in the scans, a D.C.
Multiplex lamp supply was necessary. The
scanning unit lay directly below the pro-
jectors with the phototube control unit and
phase meter directly above the lamp supply
in a cabinet. The scanning chamber could be
moved in three mutually orthogonal direc-
tions. In operation it was set in x, driven at a
constant rate in y, and controlled in z by the
phase meter. In this instrument the profile
was produced by a hot wire cutting a thin
strip of plastic set in the model carving cham-
ber, directly below the scanning unit. After
the first profile was cut, it was used as a gen-
eral guide for the scan on the second. The
second was used for the third, etc.
Looking somewhat more closely at the
phase comparison-profile traverse approach
in this instrument, we found that, primarily
due to the heavy weight of the scanning unit
and the relatively slow response in the servo
system, the possibility of obtaining a reason:
able speed of operation was limited. There-
fore, in the next instrument phase, compari-
son was abandoned in favor of amplitude com-
parison. This instrument, constructed for us
by Pickard and Burns, Incorporated, Boston,
Mass., in outward appearance is very similar
to the Bausch and Lomb instrument employ-
ing the same multiplex frame and electronic
cabinet, but its operation is quite different,
amplitude matching and a directed xy-tra-
verse. It was found that there was excessive
noise introduced by the original optical-elec-
trical transducer, and this was revised and
later replaced by a translating pinhole. Since
no profiles were being followed and since
speed was a prerequisite, the data were re-
corded on Teledeltos paper by means of cur-
rent from the matching unit. The unit would
be set in x and z and moved in y and every
time a match occurred the paper would be
marked. A change would then be made in z
equivalent to the contour interval and the
process would be repeated.
Due to a number of reasons this did not
prove to be the practical approach.
Our next effort in automatic stereoplotting
was directed toward a contour plotter which
followed the area correlation approach. This
instrument operated with a conventional
Kelsh plotter and was built by the Hycon
Company of Pasadena, California. The inten-
tion was to scan a small area, and sense the
direction of the contour. This tends to ap-
proach more closely the action taken when a
human being operates the machine. As he
scans the model, the specific point is observed,
but also the general surrounding area is taken
into consideration. Although we had high
hopes for this instrument, technical diffi-
culties kept it from being successful. The
main difficulty centered around the scanning
head; however, there were also some circuitry
problems encountered. The scanning head
has been rebuilt by Hogan Laboratories, New
York City, and we have corrected the cir-
cuitry in our Laboratories. We are continuing
investigations of this approach.
While working on the approaches just men-
tioned, The Photographic Survey Corpora-
tion of Canada was also busy in this area. Mr.
Gilbert Hobrough of that organization pro-
duced a model instrument using a Kelsh
plotter which demonstrated the feasibility of
this approach to the problem.
In place of the standard platen there was a
scanning unit 12" by 10" by 18" high, which
was mounted on ways permitting it to move
in the x, y, z directions. The illumination driv-
ing the automatic matching mechanism was
derived from a flying spot scanner, installed
with its face at the surface of the model, or
rather at the spot where the face of the plat-
ten would be. The blue light from the tube
passed back up through the projector lenses,
through the diapositive plates, and then on to
dichroic mirrors which reflected the blue light
>
J
EEE
