Slowly but surely coherent optical methods are assuming an important place in the field of optical system design, and applications of these methods are moving out of the optical research laboratories and being used and further developed by people in other fields. This seminar deals with one of those fields. It is the aim of this tutorial to provide some background into the concepts and ideas of coherent optics.
The science of photogrammetry relies on some kind of a photographic record as the source of data. Such a record represents a two-dimensional registration of, in general, a three dimensional object. After certain reduction, each photograph may be regarded as a central perspective projection of the object. This reduction is warranted because the original light ray from an object point, which is represented by a straight line in the perspective theory, is in fact a curve. In addition, the position of the latent image on the photographic emulsion undergoes shifts due to variation in environment from exposure to processing and measurement. All such effects are termed "systematic" since they can be modelled mathematically. They include: 1. Atmospheric refraction 2. Earth curvature (due to mapping a spheroidal surface on a plane, that of the photograph). 3. Lens distortion, both radial and tangential 4. Image deformation due to emulsion and base instabilities. 5. Systematic errors due to characteristics of imaging and measuring equipment.
Research on the application of coherent optical tech-niques to photogrammetric and mapping operations started during the early 60's. During the 60's most of these efforts were directed along two basic directions. On one hand various coherent optical processing techniques were being proposed and demonstrated to adlieve autocorrelation of the stereo-compilation procedure (Ref. 1) and on the other holography was being investigated for storing and displaying the information contained in the stereomodel (Refs. 2,3). In the recent past several investigations on the metric fidelity of the reconstructed holographic images have been made using conventional photogrammetric mensuratior techniques (Refs. 4, 5, 6). The potential application of holography for recording and reconstructing images for the purposes of mensuration in close range photogrammetry was the motivation behind these investigations. Results from these investigations clearly indicate that it is possible to retrieve reliable metric information about close range objects using holography. Also it has been shown that a photogrammetric quality stereomodel can be holographically stored and subsequency mapped using conventional photogram-metric mensuration techniques (Ref. 7). Holographic stereomodels of the focus image type permit reconstruction using extended white light sources. Advances made in optical memories, coherent optical pattern recognition techniques and optical detectors have all contributed to considerable progress in this area. Today applications of coherent optical techniques to mapping is an active area of research and multitude of research efforts are in progress around the country.
I am honored to have been asked to give the keynote address at this important meeting of scientists and engineers. I note with pleasure that about one-third of the papers being presented at this conference cover development activities sponsored by the Defense Mapping Agency (DMA).
Maps, representing a coding of spatial information on a geographic coordinate reference, are indispensable to many of us since we live, work and play on or very near the earth's surface. Maps serve the purpose of presenting coded terrain information of potential value to the user in his decision-making processes. There has always been a continuing need to produce accurate maps but today's increasing requirements for maps call for more rapid production of accurate maps at minimal costs. Map production involves a series of complicated processes. Future progress in producing maps faster and cheaper while retaining or improving accuracy will require either, or a combination, of two general approaches. The first and simplest approach is to improve existing technology by perfecting present equipment and techniques involved in the mapping processes. The second calls for new technologies which implies innovation and new approaches to equipment and techniques in the mapping processes. Introduction of analog and digital computers, for example, has led to sizable gains which include valuable approaches to analytical photogrammetry and automated mensuration, and present research and development indicates that many additional accomplishments will emerge. Yet there will remain requirements to improve many aspects of map production. Coherent optics has the potential for playing many roles in the mapping systems of the future.
Since 1966 the Mapping and Charting section of Rome Air Development Center has supported several research studies in the application of coherent optical processing techniques to various mapping tasks. These have primarily addressed photogrammetric operations and have demonstrated the feasibility of image matching, point transfer, parallax measurement, and image distortion detection and measurement. Several technical reports have been generated and disseminated to the DOD and civilian mapping communities. The following paragraphs provide a brief summary of the pertinent aspects of these studies with some conclusions about coherent optical processing drawn from the work accomplished.
Coherent Optics has a natural application to the processing of two-dimensional data, such as aerial photography, because it offers the ability to perform a correlation operation on large quantities of data in parallel. It is reasonable, therefore, that coherent optical systems should find application in the areas of reconnaissance and photogrammetry. We distinguish between these two areas by noting that the photogram-metrist is interested in obtaining precise metric information regarding the terrain. The reconnaissance photointerpreter, on the other hand, is often less concerned with precise measurement and more interested with determining the presence or absence of various objects of interest. For example, the photogrammetrist is interested in measuring the position of a road; the photointerpreter is often more concerned with determining the volume of traffic on the road. More-over, photogrammetric imagery is, in general, much better controlled than imagery obtained primarily for aerial reconnaissance. Consequently, while the data processing needs of these two areas are similar in many respects, there are some differences. We will, there-fore, consider separately some promising applications of coherent optics to reconnaissance and photogrammetry.
A high brightness display is achieved using tungsten sources to read out phase image plane holograms. Full color is provided by using three superimposed component holograms derived from three primary color separation object transparencies of the original subject. The holograms are contained on transparent plastic tape which is surface modulated by embossing from electro-plated masters. Any frame on the tape can be retrieved rapidly by means of the superimposed frame address Fraunhofer binary holograms which use a gallium arsenide laser to reconstruct the image on a silicon diode array. The system lends itself to post-embossing annotation. Reference Fresnel holograms recorded on each frame enable the frame to be positioned in X and Y so as to allow the map image to be driven in response to navigational tracking requirements. Because incident readout light is modulated by diffraction rather than absorption, the holograms do not rise in temperature as a result of the conversion of light to heat. This feature enables condiderably more light flux to be gated by the holograms than could be gated by any other medium of the same area. The resulting image brightness is adequate for viewing in the high ambient light environments. Since the production of color involves diffraction rather than absorption by dyed emulsion, bleaching with consequent color degradation cannot occur.
The attractive possibilities offered by volume holography, with discrimination between superimposed holograms by means of Bragg Angle selection, have stimulated much searching for photosensitive materials that permit erasable storage as well as nondestructive retrieval. The dichroism of FA centers in KC1:Na offers close to optimum optical behavior for holographic storage and retrieval, and especially a bidirectional photochromic response such that information storage without change in the overall optical density can be accomplished. High speed random access to information is a dircct consequence of the large information packing density iesulting from many superimposed holograms per unit volume. Unusually ,low noise and error rate is a unique characteristic of the particular material and storage technique employed.
Holographic techniques offer practical methods of creating automated, computer-controlled, mass storage systems for recording and retrieving cartographic and mapping information. Graphs, overlays, pictures, maps, multispectral data, etc., can be stored at high density in either digital or analog form or both on light-sensitive, nonmagnetically alterable media. Whereas conventional micrographic techniques can be used to reduce the volume of images stored using demagnification factors of 16 to 100 times or more, these techniques are limited in practice by requirements to maintain document resolution and by necessary depth of focus considerations.
Photogrammetric interest in holography dates back to 1966, with subse-quent developments to synthesize stereomodels holographically using overlapping aerial photographs. Previously reported applications of holography in mapping are reviewed. Four new general applications have been found through literature search and examination of recent research. Annotated holographic stereograms and metric quality holographic stereomodels are useful training aids for photo-interpreters and map compilers in addition to being of value to field surveyors and map users. Reproduction of high quality second generation prints from high resolu-tion first generation negatives has been demonstrated using carrier frequency contact printing. Holographic stereograms have applications in land navigation and nap-of-the earth aerial navigation. Finally, holographic stereograms are of value in advertising and recruiting people for the mapping community.
The relative ease with which three-dimensional images are reconstructed from single hologram plates makes them an attractive alternative to the photogrammetrist for mensuration and mapping of close-range objects. For accurate metric work, the reconstruction geometry must be recovered to with-in fine tolerances, the values of which are given. A simple scheme for extracting positional information from holograms, as well as performing graphic and digital mapping, is explained and sample results included. To extend the capability to topographic terrain applications from aerial photography, the concept of the Holographic Stereomodel (HS), both fresnel and focused image types, is explained and relative advantages and disadvantages enumerated. The photogrammetric and geometric problems involved in the production of HS are expounded upon. Mensuration and mapping considerations from HS together with results obtained to-date are given.
The objective of this study was to determine the feasibility of measuring terrain from aerial stereo-photos displayed in the form of holographic stereograms. Holographic stereograms are attractive for this application because they have the potential of providing a simple but vivid display with rapid access. An optical analog device was used as the basis for measurement. By using a set of Balplex (Ref. 1) plotting instruments in turn as projectors and cameras, we obtained projections of the optical terrain model and of a fabricated grid model, onto a datumplane. The left and right projections were recorded photographically and holographically, and were interpreted for measurement by an exact algorithm which is most conveniently implemented by computer.
The Fresnel holographic stereo-model described by Kurtz et al. (Ref. 1) and Balasubramanian and Stevenson (Ref. 2) provides a means of permanently storing a rectified photo-grammetric stereomodel. Briefly, the holographic stereomodel is a holo-graphic recording of a photogrammetric stereomodel such that during reconstruction the observer perceives a three-dimensional stereomodel. A complete discussion of the general characteristics of the holographic stereomodel is given by Mikhail else-where in these Proceedings.
The use of holography for the storage and retrieval of metric information in close range applications offers two distinct advantages over conventional photographic techniques. Since a hologram contains a three dimensional image, multiple perspectives are available for information retrieval. Also the hologram is capable of a larger depth of field.
Over the past several years, by many different organizations, considerable effort has been spent investigating the potential of using optical correlation as a means of detecting image registration for photogrammetric purposes. Most of these efforts, typical of basic research and development, have been carried out within limited budgets. Thus, in the interest of maximizing results, off-the-shelf components were used wherever possible, and generalpurpose mounts and fixtures were adapted for use where available.
Data reduction from stereo-photography forms a powerful tool not only for measuring the size, shape and location of an object or a scene but also to ascertain other desired parameters such as surface areas, volumes, etc. from space coordinates. In a truly vertical overlapping pair of photographs the relief information pertaining to the object or the scene is stored in terms of differential x-parallax. The x-parallax difference is the principal cause of stereoperception and it is through the measurement of x-parallax dif-ference that terrain elevation measurements are made from a pair of aerial photographs using a stereoscopic instrument. All automated stereo-compilation instruments whether electronic or optical are based on the ability to match conjugate images and automatically measure the x-parallax difference. In the electronic correlation system this is achieved by correlating the electrical signals representing the image detail generated using a scanning process. In the optical systems a direct two-dimensional correlation is performed using the intensity or amplitude distribution of light representing the image detail.
The general objectives of the COMAP study of coherent optical data reduction techniques were: 1) to provide a breadboard model and engineering evaluation of an image-matched filter correlator, and 2) to provide a design study for a large-area correlator based upon the evaluation.
Optical techniques for extracting, processing, and displaying information from recorded images have significant advantages over serial digital computers for a wide variety of applications. Of particular importance is the ability of optical systems to perform complex Fourier transform operations on large two-dimensional data arrays or images, in parallel. Widespread use of these techniques, however, hinges on the development of components for systems that utilize this parallel processing potential for high speed operation.
At present, most measurements on aerial photography are made with image-comparing instruments and shearing or filar measuring microscopes. And in most automated systems, stereo images are first digitized and then correlated with the aid of a digital computer. In all likelihood, some of these techniques can be simplified or improved by Optical Power Spectrum (OPS) techniques.
The challenge facing the mapping community today is the effective utilization of increasingly large volumes of aerial photography and other data to provide topographic information in the form of line maps, orthophotomaps, specialized products or individual responses to decision makers and other users in a timely manner. Automation of rapid screening techniques to determine the suitability of aerial photography for further processing represents an important aspect of this problem. This paper discusses the use of data from the optical power spectrum for cloud screening of aerial photography with emphasis on the computer processing of signals from a segmented spatial frequency plane detector in a coherent optical system. It is assumed that optical samples from aerial photographic patterns will have optical power spectra with attributes representitive of the patterns which can be electronically detected and classified by statistical pattern recognition techniques. This hybrid approach provides an advantageous combination of the speed of the optical Fourier transform with the flexibility of digital processing which leads to a simple, efficient system with the potential for screening aerial photography at very rapid rates.
This paper discusses two optical processing systems for the rapid extraction of contours from a pair of stereo transparencies. For a given x-parallax, these systems provide a display of all contours present at the elevation corresponding to the parallax setting.
For many years opticists have been contouring optical surfaces for deformations of the order of a few microns by the use of interferometric techniques. One of the earliest methods for doing this is to place a reference surface over the surface to be contoured. When this combination is illuminated by a coherent colli-mated beam, a fringe pattern is observed that is localized close to the surface that is being contoured. The fringe pattern (interferogram) is easily interpreted to yield contour information. This simple device (interferometer) is known as a Fizeau interferometer. (Ref. 1) As can be easily seen this experiment (and other interferometric techniques) is only applicable to surfaces which are smooth compared to the wavelength of the light used. The object must also have deformations from the reference surface not much larger than a few wavelengths of the light. In an effort to contour rougher surfaces and greater deformations than are allowed by using visible light, longer wavelength light such as the radiation from a CO2 laser is used. Unfortunately there exist many objects that are too rough and have too great a deformation to be suitable for an interferometer even with these longer wavelengths. At present, many of the objects of this type are handled by the conventional techniques of close-range photogrammetry (Ref. 2) and moire fringes. (Ref. 3)
The last decade has seen an increasing inter-est in developing alternatives to manual stereocompilation. These new methods have ranged from electronic cross-correlation to optical cross-correlation and match filtering (Ref. 1-3). This paper discusses a technique of generating a complete set of contours using a relatively simple optical apparatus. Over-lapping stereophotographs are coherently summed using a Mach-Zehnder Interferometer. For vertical photography the y-parallax can be removed by proper orientation of the film plates and the only remaining x-parallax is due to terrain elevation differences. This parallax will result in a grating-like structure that can be bandpass filtered in the Fourier Transform plane to pass some given grating spacing and the corresponding parallax. The filtered image will contain a constant parallax contour, which for vertical photography is a constant height contour. Since the grating-like structure is not sinudoidal, higher order frequencies will be present and harmonic frequencies corresponding to multiples of the parallax associated with the filter frequency will also be present and produce additional contour fringes at a reduced modulation in the image.
Vander Lugt (Ref. 1) introduced the use of holograms as frequency plane masks in spatial filtering applications. The technique produces the convolution of an input spatial scene s(x) with a convolving function a(x) by placing a hologram in the Fourier transform plane of a coherent optical processor. The hologram is a photograph of a reference beam and the Fourier trans form- H(f) of a spatial mask function h(x). The filtered scene is realized as an off-axis spatial signal. See Figure 1.