I joined the Artificial Intelligence Department of Bunker Ramo as Associate Director. Don Swanson was Director (the group changed its name to the Synthetic Intelligence Department -- SID -- to make it more attractive for me.) Their major activity was in programming a computer to translate Russian into English. IBM had a parallel contract with the Air Force and was using a brute-force approach -- they developed a photo-optical memory to handle the word-by-word translation they were employing. The SID group was using grammatical tools to help obtain the correct choice of words. One member of the group, Sergei Fomenko (born in Russia, educated in China), was also involved in developing a scanner for converting printed text into computer format, a much more interesting topic for me.
Shortly after joining the group we received a request for a proposal to automate a Kelsh Stereo Plotter. My proposal won the contract and we were in the mapping business.
The background materials for most maps are obtained from aerial photographs that cover an area of interest. The pictures are usually exposed at locations six-tenths of the aircraft altitude apart (for example, at points 6000 feet apart if the airplane flies at an altitude of 10,000 feet). This makes the triangle formed by the two positions and any ground point seen in two adjacent photos have substantial angles making it possible to measure the ground elevations to fairly high accuracy from the photographs.
The figure illustrates the photographic sequence used. The aircraft flies a straight path at a constant altitude and takes pictures at a spacing of about six-tenths of its altitude. After covering the length of the area of interest it returns on a parallel path, shifted over sufficiently to provide photographs of an adjacent strip. The process is continued until stereoscopic coverage is available for the entire area of interest
The figure shows an object on the ground that is in position to be imaged from the two exposure locations. It is imaged to the left of center in the left hand view but to the right hand center on the right hand view. This "parallax" displacement of the image between the two camera locations allows stereoscopic viewing of the pictures (our two eyes receive parallax-displaced images that our "system" interprets as differences in distance). For this type of aerial photography the separation between the two views is very large, yielding exaggerated apparent elevation changes making it possible to obtain good height measurements.
Exposure Locations for Stereoscopic Coverage of Area
Note that the image of P on either film by itself could represent any point on the respective rays, so photographs cannot make good maps. Once the elevation of points are known from the intersection of the corresponding rays, their geographic locations are also available.
Most cameras used in commercial work have a six-inch focal length and use nine-by-nine inch film. Positive prints, or "diapositives", are made on heavy glass plates from the negatives, so they are stable dimensionally and capable of yielding measurements of high accuracy.
Many commercial companies use such diapositives for mapping operations. They commonly use a "Kelsh Plotter" to make the detailed measurements. It consists of two projectors that are placed on a bar and adjusted so that they model the camera locations at the times of exposure. Small sections of the two diapositives are projected onto a small matte-surfaced "platen" of about two-inches in diameter that is carried on a precision vertical screw, so it can be moved to known heights above the table. The "platten" has a small point of light at its center to mark the point of interest. One image is projected onto the platten through a blue-green filter and the other through a red filter and the operator wears a pair of glasses with corresponding filters that steer the images appropriately to the two eyes, so a stereoscopic view of the small areas projected on the platten is obtained. The operator might adjust the height of the platten to a height above the table corresponding to an altitude of interest and move the "platten" around to locate the "contour lines" of that elevation. These are traced on a map sheet placed on the table.
Before any measurements can be made the projectors must be adjusted to correspond, very exactly, to the elevations (to the desired map scale) and orientations of the camera at the time the two pictures were taken. This is done by adjusting the projectors so as to place images having known locations in proper relationship to the table, a very demanding and tedious operation.
For automatic operations with the Kelsh Plotter, the platten was replaced by a Nipkow-Disc Scanner. This device, the original TV scanner invented in 1885, consists of a rotating wheel with a set of holes that move, one by one, across the area of the "window" formed by the projected rays. The rays corresponding to transparent areas on the films, passed through to corresponding photocells below when a disc opening passed them. The many holes in the disc were placed on a spiral so that they scanned across the window area, in a crude TV-like manner, on each revolution. As described in Appendix 8, this provided the signals for the device to keep on the coincidence of the rays as the scanner moved across the field. As the device moved across the field the altitudes were printed on a film and the images on the film, as seen by the scanner, were used to expose a new film in which each elementary image appeared at its proper location on the evolving map; these orthographic projection photographs, or "orthophotographs" provided the background for a new type of topographic map.
Our customer supplied us with the Kelsh Plotter along with a hay-wired scanning system prepared by an earlier contractor who had not made it work. We soon had it working (sort of, it was too flimsy to do a good job). and went on to build a finished unit. It printed out the measured altitudes as a sequence of three color bands and, concurrently, produced an orthophotograph. This work was covered by a first mapping patent 9.
Fig. 2 is a photograph of the "Automatic Stereomapper" and Figures 3 and 4 are copies of an altitude chart and orthophoto made by it.
The speed of operation of the unit was limited by the ability of the servo motor to move the scanner up and down. Some time after proceeding on the computerized equipment, described next, it became evident that the performance of the Stereomapper could have been substantially increased by making some altitude adjustment independently of the mechanical height changes. This was covered by another patent 10, but never implemented.
In this period, I did some consulting with TRW with respect to tracking the ground under an airplane to enable the taking of sharper photos. For the purpose I proposed the use of a Nipkow Disc, with a magnetic track on the disc to store the ground image, as a reference for the tracking. This was demonstrated and is covered by a patent 11.
About this time I also had the idea that the general mapping scheme would be useful in obtaining detailed surface data from small physical models. This resulted in another patent 12, the only one I have had contested. It seems someone at Boeing had a similar idea and they tried to show that theirs was first but mine was upheld. The idea was never used at Bunker Ramo, but I suspect it was used at Boeing.
While this work was progressing we received another request for proposal, from the same customer, for an "Automatic Map Compilation System" (AMCS), another device for automatic mapping. It was to use a computer to set up the stereo model instead of the Kelsh Plotter. My proposal again won and we received a second contract. By about 1962 both units were operating and the customer was emboldened to go out for bid on a prototype unit for production operations, the UNAMACE (Universal Automatic Map Compilation Equipment); we won it too.
The AMCS incorporated a Packard Bell 250 computer. This served us well, but gave our customer problems after we delivered it. It had a delay line random access memory that was temperature sensitive, so when the customer turned off their air conditioning at night the memory wouldn't work in the morning and they would adjust the delays. But they would also start the air conditioning, so before long the memory was again out of adjustment! The UNAMACE used a Bunker Ramo computer designed for submarine service, so it sold a few of the companies computers, but it was far from an ideal choice for the purpose; whereas the positions of the tables were adjusted to two microns and the long axes of the tables were 457,000 microns, so eighteen bits were required to keep track of the table positions the computer had a fifteen bit word. In addition the digital to analog converters of that day had only a ten bit accuracy -- I suspect that my solution to this problem is what kept the equipment confidential long after the usual period.
I laid out the detailed flow charts for the computer programs and, on the AMCS a good programmer carried out the detailed programs -- but I did write some auxiliary programs. We started out with a professional programmer on the UNAMACE also, but it didn't go well. Much of the early programming was discarded and I did a lot of it over with the help of an Army Map Service employee who was assigned to work with us to learn the equipment.
The operations of the AMCS and UNAMACE were similar, so they are described below using the UNAMACE configuration.
The UNAMCE employed four identical high precision tables, each capable of carrying diapositives of up to 9"X18" (quarter inch glass plates, so they are quite heavy) or unexposed film of the same dimensions. The measuring units provided counting signals at two micrometer 80 millionths of an inch) intervals that stepped counters with associated digital to analog (D/A) converters. The computer output to corresponding D/A's and the differences between their outputs, showing the error in the table positions from the commanded positions, were amplified to drive the tables.
Flying spot scanners (FSS -- high intensity TV-like tubes), with associated optical systems, replaced the Nipkow-Disc of the Automatic Stereomapper. These were used to scan the diapositives on two of the tables and to expose the films while the other two tables to prepare the orthophotos and altitude charts.
The weight of the tables plus diapositives led to considerable sluggishness in their accurate positioning. For this reason, the errors in the table positions from their commanded positions was input to the deflection circuitry of the FSS and adjusted to achieve "stop motion" where the scanning centered on the computer-indicated positions on the photographs even when there was substantial error in the table positions. This permitted an essentially instantaneous response to computer commands, an important element in speeding up the operation. A major breakthrough with the AMCS came on the day the customer was to see the equipment in operation for the first time. I was called in because things weren't working and found that the development team had not adjusted the stop motion properly. It was easy to do -- pushing on the tables should not have moved the images on the viewer and they were adjusted so it didn't.
The operator used a stereo viewer, with an electronically generated cross-hair, to assist in the operations. In the AMCS this took the form of a pair of two-inch diameter cathode ray tubes (CRTs); at the start, the operator viewed these with a simple stereo viewer (short focal length lenses with prisms), but the glasses were not used for long as all involved soon learned to view the left CRT with the left eye and the right CRT with the right eye without it. Four-inch CRTs were used in the UNAMACE, so a viewer was required for it. It took the form diagrammed below.
UNAMACE Stereo Viewer
The two CRTs were stacked vertically. The lower CRT was viewed through a half-silvered mirror (a mirror that passes half the light and reflects the rest) and the upper via reflections from the two mirrors (the lower CRT was moved back to make the path lengths equal). Polaroid filters in the two light paths and corresponding glasses worn by the operators were used to separate the images to the two eyes. The CRTs produced TV-like reproductions of the small diapositive areas under examination at the time.
Before "compilations" (height measuring operations) the photographs were subjected to a "comparator" operation in which the photo coordinates of a number of easily identified points in the photographs were measured. Many of these points would have known map coordinates from earlier ground survey work, so a computer calculation could establish the relationship between the ground coordinates and the photo coordinates. While the measurements for this operation were performed using the UNAMACE during our tests, after delivery to the customer the measurements and calculations were made on other equipments.
With the diapositives mounted in the equipment, the computer would command the tables to move to the vicinity of "fiducial marks", placed on the edges of the pictures by the camera that had been used as references during the comparator operation. The operator, through commands to the computer, centered these accurately in the fields of view, thus relating the table coordinates to the photo coordinates and making it possible for the computer to locate points in the field using their geographic coordinates.
The computer would next move the tables to the starting location for the compilation (a corner of the rectangular terrain area to be mapped from the particular stereo pair). As the altitude of this point is generally not known, the computer used a guessed altitude for the purpose and then waited for the operator to adjust the altitude as indicated by the cross hair in the stereo viewer. With this accomplished the computer began to move the scan in small steps, determining the correct altitude at each step and printing out the altitude and orthophoto element at each step. (Later the customer found it more useful to store the measured altitudes digitally and prepare the charts on simpler equipments. New electronics were supplied for the tables left over to make them into new height measuring systems).
During the acceptance tests we demonstrated that the equipment could reduce the size of the outputs and combine the number of stereo pair outputs required to form a complete map sheet. I also devised a test model that used a pair of diapositives made from the same negative, with different data input to the computer for the two so the model appeared like a plane with a very slight tilt (and some curvature because the equipment compensated for the curvature of the earth -- and I forgot to disable the feature.) This model provided a very sensitive test of the equipment accuracy since the data used and results to be met didn't come from other sources.
I had given a number of talks before the American Society of Photogrammetry on the earlier units and, in 1964, with the blessings of the contracting agency, I gave a talk in Lisbon, Portugal on the features of UNAMACE. In 1968, again with the blessings of the agency, I gave a talk telling of its operational characteristics in Switzerland. After that I was told not to hand out the two papers together although our enemies were at both conferences! When the equipment became it proved to be far better than the government people had anticipated so they had decided that many of its features should be classified. Despite my talks in Europe, a number of the features of UNAMACE remained classified for about 20 years (confidential items are normally declassified within a few years). Since the customer had considered it unclassified at first I surmise that some details were being used in the Star Wars program.
The UNAMACE of that period measured ground elevations from standard aerial photographs to about 1/10,000 of the altitude that the photographs were taken from (one foot in ground elevation error from photographs taken from a 10,000 foot altitude) and could do it at about 50 measurements per second. The computer program allowed the operator, before compilations started, to designate areas in the photographs that were likely to be troublesome; for example, waves on water would supply no useful information so the operator would trace around water areas ahead of time and instruct the computer to just hold the altitude determined for the boundary until it reached the other side. The computer permitted operations with unconventional, including quite (but predictably) distorted types of photography that our intelligence people were interested in.
The preparation of the diapositive data required for compilations interested me. I learned to solve multi-coordinate non-linear equations using least squares and prepared "relative orientation" programs that permitted a model to be set up without any ground data, as would be required for moon mapping, for example, by correlating the measurements of images of common terrain points on two photographs. We also prepared "absolute orientation" programs that used ground survey data to determine the photographic parameters.
Also, the accuracies required made it necessary to correct for the fact that light rays do not travel along straight lines through the atmosphere, so I became an "expert" in this area and had a number of papers on the subject published 13.
In 1969 I received the Photogrammetric Award of the American Society of Photogrammetry for the development of these equipments. Some time earlier the customer personnel had been given an award, in a television presentation in Washington, for their part in the development.
I had one more patent granted in this period. It was to use a Flying Spot Scanner (FSS) to expose relatively slow photographic film. The problem had been posed to me during a discussion and I had suggested that the film should be moved around, with the FSS beam following, so the beam wouldn't have to stay stationary on the FSS to a point where it would burn its face. The patent is listed under S. Bertram et al 14, the et al being the man I had been talking to. He told me afterwards that he was certain I wouldn't apply for a patent on the idea, so he did it in both of our names to get it on the record.
On to How Not to Run a Railroad
Top of Page
Back to Table of Contents
Back to Sidney Bertram
Back to Jacques & Anne
Back to The Abramovich & Wilder Families
Back to Bertram Home Page
Master Index