At 10:30 a.m. EST on January 28, 1986, 73 seconds after liftoff from its Cape Canaveral launch pad, the space shuttle Challenger was destroyed by a catastrophic explosion and the seven crew members aboard were killed. I wish to dedicate this editorial to the memory of her crew: Francis R. Scobee-Shuttle Commander, Michael J. Smith-Shuttle Pilot, Ronald E. McNair-Mission Specialist, Ellison S. Onizuka-Mission Specialist, Judith A. Resnik-Mission Specialist, Gregory B. Jarvis-Payload Specialist, Christa McAuliffe-Space Flight Participant. I believe that I am speaking for all the officers, governors, members, and staff of SPIE in expressing our heartfelt sorrow to the families and friends of these seven dedicated space pioneers. I also trust that I am speaking for all of us in SPIE in encouraging the United States to press ahead with its space program and to ensure that the Challenger's crew did not give their lives in vain.
Spectrometric Techniques, Vol. IV-Reviewed by Hermann Robl; Flow VisualizationIII-Reviewed by Bruce V. Johnson; Industrial Applications of Lasers-Reviewed by Francois M. Mottier; Practical Laser Safety-Reviewed by Jerry J. Anderson
Intelligence evolves out of matter, so said the Sankhya philosophers of ancient India. The discipline of artificial intelligence (Al), which was established some 30 years ago, has confirmed the validity of the above assertion. Recently, a number of AI applications have been successfully demonstrated, generating a great deal of excitement and interest in scientific and technical circles. In this special issue of Optical Engineering a representative set of applications that incorporate Al principles is presented.
This paper describes a flexible stereo verification system, STEREOSYS, and its application to the analysis of high resolution aerial photography. Stereo verification refers to the verification of hypotheses about a scene by stereo analysis of the scene. Unlike stereo interpretation, stereo verification requires only coarse indications of three-dimensional structure. In the case of aerial photography, this means coarse indications of the heights of objects above their surroundings. This requirement, together with requirements for robustness and for dense height measurements, shapes the decision about the stereo system to use. This paper discusses these design issues and describes an implementation.
In this paper we examine a number of issues related to developing a vision system for aerial scenes. A hierarchical vision system that utilizes evidence from intrinsic and relational characterizations of objects is described. Issues related to developing spectral and spatial operators for extracting image cues about the presence of objects are discussed. Example analyses of high resolution urban scenes are presented.
Human photointerpreters use expert knowledge and contextual information to help them analyze a scene. We have experimented with the Lockheed Expert System (LES) to see if contextual information can be useful in interpreting aerial photographs. First, the gray-scale image is segmented into uniform or slowly varying intensity regions or contiguous textured regions using an edge-based segmentation technique. Next, the system computes a set of attributes for each region. Some of these attributes are based on local properties of that region only (e.g., area, average intensity, texture strength, etc.); others are based on contextual or global information (e.g., adjacent regions and nearby regions). Finally, LES is given the task of classifying all the regions using the attribute values. It utilizes multiple goals and multiple rule sets to determine the best classification; regions that do not satisfy any of the rules are left unclassified. The authors obtained the rules by an introspection technique after studying many aerial photographs. Unlike programs that use only statistics in the region under consideration, LES can use contextual information such as the fact that cars are likely to be adjacent to roads, which significantly improves its performance on regions that are difficult to classify.
It has been argued that knowledge-based systems (KBSs) must reason from evidential information, i.e., from information that is to some degree uncertain, imprecise, and occasionally inaccurate. This is no less true of KBSs that operate in the domain of computer-based image interpretation. Recent research has suggested that the work of Dempster and Shafer (DS) provides a viable alternative to Bayesian-based techniques for reasoning from evidential information. In this paper, we discuss some differences between the DS theory and some popular Bayesian-based approaches to effecting the reasoning task. We then discuss some work on integrating the DS theory into a knowledge-based high-level computer vision system in order to examine various aspects of this new technology that have not been explored to date. Results from a large number of image interpretation experiments are presented. These results suggest that a KBS's performance improves substantially when it exploits various features of the DS theory that are not readily available in pure Bayesian-based approaches.
Multiple sensing is the ability to sense the environment with the concurrent use of several sensors. There are currently a number of different sensors routinely used in image processing applications, and the trend is toward the development of more sophisticated and less expensive sensors. This trend is complemented by the development of parallel and multiprocessor architectures for processing the large amounts of data collected by these sensors. The capabilities of many image processing systems can be greatly enhanced by the organized use of several types of sensors and by the develop-ment of methods capable of integrating the collected data in a way that can yield information otherwise unavailable, or hard to obtain, from any single type of sensor. The advantage in using several similar sensors has already been demonstrated in the context of dynamic scene analysis, where information contained in a sequence of visual intensity images (including stereoscopic images) has been integrated with the twofold objective of obtaining the three-dimensional description and the motion of objects in space. The advantage in using several different sensors is clear from the quite obvious observation that different sensors are sensitive to different signals, each one of which can reveal a particular set of properties of the sensed environment. This paper discusses the advantages of multiple sensor integration/fusion with different sensors through image processing and identifies a number of associated problems. It reviews preliminary work on the solution of these problems and indicates the direction of future research.
Many problems in motion can be approached by focusing attention on events in the motion of objects. Our concept of event is the very general one of any discontinuity in consistent motion. This paper addresses the use of low-level events comprising discontinuities in the regular consistent motions of feature points. For the experiments described here, events are changes in the parameters of uniformly accelerated motion: acceleration, velocity, and initial position. We then use the detected events and the concept of path coherence to achieve a correspondence that describes the motion of objects over many frames. The correspondence mechanism seeks to minimize the number of events.
Three-dimensional object shape and orientation are determined from a single perspective view with very little a priori knowledge. Groups of parallel lines are detected by searching for common vanishing points, and the orientation of a group of parallel lines is determined by the location of its vanishing point. Edge contour traces are made for finding a primary face of the object. A least-slant-angle heuristic is adopted to help find the orientation of the primary face. A model of the 3D object is constructed in steps, beginning with the primary face. The shape and orientation of the faces in the model are adjusted iteratively so as to minimize the discrepancies between the object image and the image generated by the model. Three examples are presented, and the resulting 3D model interpretation of the object is consistent with human perception.
A vision system is described for depalletizing steel cylindrical billets for a forging application. An algorithm for accurately locating and measuring the billets is described in some detail. Highlights of this discussion include an algorithm for adaptive threshold selection to accommodate changing image brightnesses and a special robot calibration procedure that enables inference of depth using only a single camera view combined with prior knowledge about the scene. Experimental results show that the system provides accurate measurements in spite of poor, inconsistent contrast.
The Computer Vision Laboratory at the University of Maryland for the past year has been developing a computer vision system for autonomous ground navigation of roads and road networks for the Defense Advanced Research Projects Agency's Strategic Computing Program. The complete system runs on a VAX 11/785, but certain parts of it have been reimplemented on a VICOM image processing system for experimentation on an autonomous vehicle built for the Martin Marietta Corp., Aerospace Division, in Denver, Colorado. We give a brief overview of the principal software components of the system and then describe the VICOM implementation in detail.
The advent of advanced computer architectures for parallel and symbolic processing has evolved technology to the point at which prototype autonomous vehicles are being developed. Control of such devices requires communication between knowledge-based subsystems in charge of the vision, planning, and control aspects necessary to make autonomous systems func-tional in a real-world environment. The performance of autonomous vehicle systems is currently limited by their inability to accurately analyze their surrounding environment. In order to function in dynamic situations, autonomous vehicles must be capable of interpreting terrain on the basis of predetermined mission goals. This paper describes an autonomous airborne-vehicle simulation currently being developed at the Georgia Tech Research Institute. The Autonomous Helicopter System (AHS) is a multimission system consisting of three distinct sections: vision, planning, and control. The vision section provides the local and global scene analyses that are symbolically represented and passed to the planning section as the initial route-planning constraints. The planning section generates a task-dependent path for the vehicle to traverse that assures maximum mission system successes as well as survivability. The control section validates the path and either executes the given route or feeds back to previous sections in order to resolve conflicts.
This paper presents a comprehensive approach to the design of machine vision software for roving robots and autonomous vehicles. Various techniques are proposed for solving the important problems of directional guidance, obstacle avoidance, and object identification. Artificial intelligence and knowledge-base concepts form the basis of the vision system design. The principle of texture invariance is introduced for shadow analysis and discrimination. The idea of scene layout footprints and 3-D maps for landmarks is proposed as a means of orientation determination for the guidance and navigation of roving robots and autonomous vehicles. The vision system performs three phases of visual processing: the initialization phase, the "walking" phase, and the warning phase. The visual processing and interpretation are monitored by the knowledge access and inference routine.
The original expert systems for the most part were handcrafted directly, using various dialects of the LISP programming language. The inference and knowledge representation components of these systems can be separated from the domain-specific portion of the expert system and can be used again for an entirely different task. Some of these tools, generically called shells, are discussed. Although these shells provide help in building knowledge-based systems, considerable skill in artificial intelligence programming is still necessary to create an expert system that accomplishes a nontrivial task.
This paper presents the design and development of a prototype document retrieval system using a knowledge-based systems approach. Both the domain-specific knowledge base and the inferencing schemes are based on a fuzzy set theoretic framework. A query in natural language represents a request to retrieve a relevant subset of documents from a document base. Such a query, which can include both fuzzy terms and fuzzy relational operators, is converted into an unambiguous intermediate form by a natural language interface. Concepts that describe domain topics and the relationships between concepts, such as the synonym relation and the implication relation between a general concept and more specific concepts, have been captured in a knowledge base. The knowledge base enables the system to emulate the reasoning process followed by an expert, such as a librarian, in understanding and reformulating user queries. The retrieval mechanism processes the query in two steps. First it produces a pruned list of documents pertinent to the query. Second, it uses an evidence combination scheme to compute a degree of support between the query and individual documents produced in step one. The front-end component of the system then presents a set of document citations to the user in ranked order as an answer to the information request.
The architecture of the control and data structures of the real-time expert system ERIK ("evaluating reports using integrated knowledge") is fine tuned for the task of flexible knowledge-intensive interpretation of ship messages. In spite of the large amounts of knowledge needed for this interpretation task, ERIK can interpret noisy and ill-formed inputs with high reliability due to a variety of computational features. Among these features are an expectation-driven control structure, selective variable depth focusing of attention, intelligent recovery from failed expectations, distributed recognition modules, and a sophisticated spelling correction facility. ERIK, which is now running on-line at the Coast Guard Automated Mutual-assistance Vessel Rescue (AMVER) center, incorporates artificial intelligence design and programming techniques. Using these techniques to solve a complex real-time information processing problem has led to theoretical and practical insights that are presented in this paper.
We have improved the performance of the avalanche-transistor deflector-driver (sweep) circuitry used in the high-speed, electro-optic streak camera at Lawrence Livermore National Laboratory. In the previous design for the sweep circuit, trigger-to-output delay time drifted on some cameras. This delay drift is a function of a somewhat randomly unstable breakdown voltage of some avalanche transistors. Both temperature and differences in manufacturing methods for transistors affect this instability. However, a significant improvement in system performance is achieved by long-term burn in and by selection of only the most stable transistors for the sweep circuit. The peak-to-peak sweep voltage has been increased about 80% by increasing the number of avalanche transistors in the string and raising the quality factor (Q) of the resonant circuit. The result is an improvement in sweep uniformity by a factor of approximately 2. Design equations for selecting components are given. Fast-recovery diodes are used to prevent undershoot and to keep the beam out of the intensifier field of view until after the intensifier is gated off. The sweep time range has been extended to over 100 ns.
A study of the spectral nature of the diffuse attenuation coefficient of light, K(X), for various types of oceanic waters has been performed. These attenuation spectra were computed from downwelling spectral irradiance data Ed(X) obtained by U.S., French, and Japanese in-vestigators working in widely separated oceanic regions and using different measuring techniques and equipment. Attenuation properties were calculated over the spectral region from 365 to 700 nm and for depths from near-surface to in excess of 100 m. Examining the K(X) data, we find that strong, simple, and useful relationships exist between the value of K at some selected reference wavelength, Xo, and the value of K at some other wavelength such that K(X) = M(X) [K(X0)-Kw(X0)] + Kw(X), where Kw is the attenuation coefficient for pure sea water. For oceanic waters (for example, Jerlov types I through III), the relationships are linear. These relationships appear to be useful throughout the entire spectral range examined and are particularly good between 420 and, say, 580 nm. The significance of the existence of such relationships is that they allow the inference of the spectral attenuation coefficient at all wavelengths from the attenuation value at a single wavelength and also provide analytical expressions for modeling the spectral nature of the attenuation in ocean and clear coastal water.
The design and performance characteristics of a compact, wide-band, space-integrating, acousto-optic signal processing system that provides the unipolar convolution of high-speed digital pulse trains is de-scribed. The system utilizes a semiconductor diode laser, two tellurium dioxide (Te02) Bragg cells with bandwidths of 100 MHz and aperture times of 6 pis, imaging and collection optics, and an avalanche photodiode to perform the convolution operation in real time.
It is shown that a mode-locked laser can be used as a local oscillator (LO) in heterodyne ranging to bandwidth compress the received signal. The amount of bandwidth compression depends on the difference in the intermode frequencies between the LO laser and the transmitter laser, which can also be a mode-locked laser. Since the heterodyning occurs on the detector, it is only necessary to use a narrow bandwidth receiver that passes the compressed bandwidth to obtain ranging equivalent to that obtained by a wide bandwidth receiver that passes the full transmitter bandwidth. Bandwidth compression of over 500x and 1 ft range resolution was experimentally demonstrated when a 1.5 GHz He-Ne laser mode-locked transmitter signal was reduced to less than 3 MHz at detection.
A simple noniterative line-thinning algorithm, which is suitable for nearly straight interferometric and moire fringes, is proposed. The algorithm was found to be fast and accurate. An error analysis is presented.
A versatile and useful personal computer based procedure to generate halftone representation of digitized image data is presented. This feature provides limited interactive image processing capability for a basic personal computer work station without having to acquire a video monitor and a digital-to-analog converter board. The generated halftone images can be displayed on the standard dot-addressable graphics monitor, and/or they can be printed on regular paper via an associated printer. The halftoning procedure is based on an error diffusion technique in which dots are produced randomly, one at a time, based on both the local gray level value and some accumulated measure of the error due to previous assignments. The key elements in halftoning, such as dynamic, spatial, and display medium resolutions, are defined, and the mathematical correspondence among them is established.
Calculations are made of regression coefficients in both the wavelength and spatial frequency domains for relative humidity, air temperature, and wind speed, with respect to atmospheric contrasts as measured over a three-year period. Significant changes are noted between summer and winter, including some sign changes and opposing wavelength dependences. Analysis of spatial frequency data permits determination of the effects of each meteorological parameter on background, turbulence, and aerosol modulation contrast functions separately. Results indicate that in the rainy season, when the atmosphere is freer of airborne soil-derived particulates, turbulence is dominant in limiting imaging resolution through the atmosphere, with wavelength dependence determined primarily by background and for ward scattering effects associated with humidity. Resolution is best in the near-infrared. However, in the dry season image quality is limited primarily by large airborne particulates and their effects on atmospheric background and spatial frequency dependent multiple forward scattering phenomena. In this case, resolution is best at short wavelengths. The strong wavelength dependences on small and large radii aerosol related effects suggest the possibility of predicting imaging resolution spectral dependence in advance in accordance with meteorological predictions. Analysis of regression coefficients in the spatial frequency domain permits quantitative determination of the effects of each meteorological parameter on each type of atmospheric MTF, i.e., background, aerosol, and turbulence MTFs, separately. In this way insight is gained not only as to the extent to which each meteorological parameter affects imaging resolution but also as to the basic mechanism of the effect.