Advanced version of MINACE (minimum noise and correlation plane energy) filters are applied to detection. Detection involves locating all objects in multiple classes in an image independent of the object class, distortions and contrast differences, i.e. we do not attempt to recognize the object class. This is the first step in scene analysis. A number of false alarms are expected per scene and will be reduced in subsequent levels of the full processor. The advanced MINACE filter concepts we use involve: including a constant background in the training images, training on intermediate contrast ratio objects, use of a new background clutter model and use of orthogonal filter sets.

Conventional synthetic discriminant function (SDF) filter designs use all available training images by requiring that the designed SDF filter yield prespecified values at the origin of the correlation plane when various training images are used in the input. In this paper, we show that we can reduce the number of training images being used and improve the filter performance by relaxing the correlation plane constraints when threshold detection is employed in the output correlation plane.

A general class of distortion invariant quadratic filter is developed for optical pattern recognition. It produces an equal response to a set of training images while maximizing the normalized response. Different forms of the filter are obtained by constraining the filter to operate on different linear subspaces of the input signal. These subspaces can be used to represent constraints due to system components or they can represent feature spaces which model the internal structure of certain types of signals. The proposed quadratic filter can be mapped onto a variety of different optical processor architecwres providing a general framework for developing optical processors for pattern classification.

Any desired diffraction pattern can be produced in the Fourier plane by specification of a corresponding input plane transparency. Complex-valued transmittance is generally required, but in practice, phase-only transmittance is used. Many design procedures use numerically intensive, constrained optimization. We, instead, use a noniterative procedure that directly translates the desired, but unavailable, complex transparency into an appropriate phase transparency. At each pixel the value of phase is pseudorandomly selected from a random distribution whose standard deviation is specified by the desired amplitude. We apply the pseudorandom phase-only encoding to hybrid composite filter design. These filters are used in a filter bank architecture to perform intensity- and distortion-invariant pattern recognition.

An experimentally based technique is presented which improves the signal to noise ratio in the detector plane of an optical correlator. The technique utilizes off-line processing by a digital computer at the current stage, but optical and hybrid architectures can be envisioned which would allow the processing to be done in real time (speed of detector).

An electro-optic processor (EOP) incorporating a miniature ruggedized optical correlator (MROC) has been fabricated for use on a remotely piloted vehicle (RPV). The EOP consists of a single-board computer for system control, a MaxVideo 20 card for interfacing to the sensor and performing image processing functions, and an MROC module. The MROC and associated electronics (SLM drive electronics, CCD readout electronics, laser controller, preprocessor, and controller) are configured in a chassis that is placed into an RPV with a visible camera for signal input and a telemetry system for output of the optical processor to the ground.

In 1986, the U.S. Army Missile Command (MICOM) unveiled the world's smallest optical correlator to demonstrate the ability to make small rugged optical processors capable of surviving transportation and use outside the laboratory. That capability has now been extended. A new solid block correlator has been built for MICOM by the Optical Corporation of America Applied Optics Division. This advanced solid block correlator (ASBC) is based on the same designed used for the original solid block correlator, but incorporates liquid crystal televisions in both the input and filter planes. The ASBC also has the advantage of two separate filter channels to increase filter throughput or allow simultaneous monitoring to the response to two different filters. This paper will present the results obtained from evaluation tests performed with the ASBC.

Experimental correlation results using invariant filters in an acousto-optic processor are presented and compared with simulation results. Zero and 15 phase state filters are used. Phase-encoded input as well as intensity input examples are given. A variety of invariance ranges are used, with target-specific discrimination included in the filter construction. The acousto-optic processor is described.

The reflected mode magneto-optic spatial light modulator (R-MOSLM) has been developed over the past couple of years. This development has led to a device that has state-of-the-art performance and is producible. This SLM device is truly compatible with semiconductor manufacturing techniques and is now being fabricated in a production environment. Performance details of individual devices is presented elsewhere. However, in this paper we discuss the measured parameters of multiple devices for statistics, discuss yield and packaging, and describe the impact of its manufacturability on cost. The system description of the correlator system using these devices is reported in a companion paper.

A new drive circuit for the liquid crystal display (LCD) of the InFocus TVT-6000^TM video projector is currently under development at the U.S. Army Missile Command. The new circuit will allow individual pixel control of the LCD. This paper will discuss results of the effort to date.

We consider distortion-invariant filters for detection (i.e. to locate a number of different object classes). For each object, there are two different depression angles, four different contrast ratios, and 18 different aspect views. The objects are present in a variety of different real background clutter. One filer is able to recognize (detect) all 2 X 4 X 18 X 5 equals 720 object versions in clutter with no false alarms using N_T equals 36 training set images. The filter uses training objects in a constant background, correlation peak constraints on the N_T objects, and minimizes a weighted combination of the correlation plane energy due to the distortion spectrum and a noise spectrum. The new object and noise models used produce this excellent performance with no false class clutter training.

Because of their suitability to two-level spatial light modulators, binary phase-only synthetic discriminant function (BPOSDF) filters are of much interest in optical pattern recognition. In this paper, we will compare three BPOSDF filter design methods using a common database. The three methods are Relaxation Algorithm, Successive Forcing Algorithm and a Phase Grating Approach recently introduced by us. The figures of merit used for this comparison include: Fisher Ratio, Recognition Rate, and Light Efficiency.

Tracking algorithms that follow a correlation peak benefit from a close relationship between center of target and center of correlation peak. Filters constructed on spatial light modulators with coupled phase and amplitude may be optimized according to various criteria. Matched- phase filters are guaranteed to center the correlation peak, but matched phase does not result from optimizing a wide variety of metrics on a coupled modulator. Laboratory results reported here bear out previous conjecture that the largest peak shift is less than half a pixel per axis.

Synthetic estimation filters (SEFs) have been found useful for the determination of distortion parameters, such as angle of rotation of objects from their images. Previous SEFs, designed using matched and phase only filters required a knowledge of exact location of the object to produce a reasonable accuracy of estimation. To overcome the limitation, use of MACE filters is proposed for the construction of synthetic estimation filters. As the MACE filters are designed to have their peaks at the origin of the correlation plane, detection of the peak can be done without a knowledge of the exact position of the object. Computer simulations are used to study the performance of MACE-SEFs. It is found that significant errors result at rotation angle which are not part of the training set. An improved linearity leading to an increase in accuracy was realized when the MINCE concept was used in place of the MACE concept for the design of the SEFs.

Several important advances have been recently made in the design of correlation filters. Among these, the minimum average correlation energy (MACE) filter is designed to yield sharp easily distinguished correlation peaks. To achieve this, the MACE filter minimizes the energy of the correlator output. We now introduce a windowing technique to control the distribution of the energy to further improve the correlation plane structure. The windows can be used to enhance the quality of the peak and reduce sidelobes by mobilizing the energy to unimportant regions of the correlation functions. This technique is particularly well suited for 1D signal processors such as acousto-optical correlators.

Phase-modulating devices, especially spatial light modulators, are often incapable of producing a full 360 degrees of phase modulation. Other limitations due to calibration errors, signal distortion, and quantization can cause the actual phase modulation to differ from the desired modulation. Such limitations on the filter plane modulator can reduce the performance of phase-only correlators. We quantify these performance losses for various phase limitations, both through simulation and through the development of an approximate model of performance. In one case we quantitatively compare the performance of phase-only filters that are optimized for limited-range phase modulation (as prescribed by Juday's `minimum Euclidean distance' principle) with the performance of nonoptimal filters. In another case we analyze the effect of not compensating a quadratic dependence of phase on signal voltage which is anticipated for some spatial modulators, e.g. deformable mirror devices.

Spatial coherence in optical processing can be exploited to implement a wide variety of image processing functions. While fully coherent systems tend to receive the most attention, spatially noncoherent systems can often provide equivalent functionality while offering significant advantages over coherent systems with regard to noise performance and system robustness. The term noncoherent includes both partially coherent and fully incoherent illumination. In addition to the noise immunity advantage, noncoherent diffraction-based processors have relaxed requirements on pupil plane spatial light modulator characteristics. In this paper we provide a discussion of the tradeoffs between coherent and noncoherent processing, taking into account the limited performance characteristics of commercially available spatial light modulators. The advantages of noncoherent processing are illustrated with numerical and experimental results corresponding to three different noncoherent architectures.

The phase-modulating capability of a commercial LCTV has been measured and its application in a joint transform correlator, using a single spatial light modulator, is discussed. The maximum LCTV phase depth obtained is closed to (pi) /2 and a small cross-coupling amplitude modulation is observed. Simulation and experimental results are obtained for the phase joint transform correlator, where the wavefront error, due to the LCTV, is corrected with a phase conjugation technic.

A new method is proposed to correct the nonlinear distortion of a practical hardware optical implementation system. The on-line optical-computer feedback technique and off-line digital iterative algorithm are used to correct the correlation matrix of a SDF and the input-output vector respectively. As a result, a optimum SDF with very good correlation S/N, minimum peak variance, is obtained. Experiments demonstrate that this method is very effective for correcting the nonlinear distortion of a real-world nonideal optical and optoelectric system, so it could be applied to 3D distortion invariant real-time optical correlation filtering and optical pattern recognition.

This paper presents a simplified description of coherent and noncoherent optical correlators and derives the correlator models. The advantages of the tolerances in component positioning and quality for the noncoherent correlator are explained in detail. Performance measures for correlator comparison are outlined and quantitative comparison data are provided. An implementation approach for bipolar noncoherent filters is introduced and promising results simulating a limited number of amplitude and phase levels are given. We will show that the performance of noncoherent and coherent correlators are comparable and thus the noncoherent correlator should be more widely used because of its implementation advantages.

This paper concerns the realization of a noncoherent correlator: the positioning tolerances and light budgets of a noncoherent correlator, the effect of the wavelength spread of the noncoherent source, the need for multiplexing noncoherent filters, the best input device (CRT or SLM) to use in a noncoherent correlator, and the light budget for the different noncoherent correlators. We conclude that a laser diode with a diffuser and input SLM laser diode array is the best choice for light source and input device. We demonstrate that a laser diode array can also be used to multiplex CGH's and that this is the preferable architecture.

We compress the range of the gray level to obtain the phase modulation from 0 to 2 (pi) . We generate the kinoform using the iterative method that includes the phase and the cross couple amplitude modulation characteristics of the LCTV. As a result we obtain good quality on-axis reconstructed images. We find the random phase errors on the LCTV pixels that causes speckle noise on the reconstructed images. Optical experimental results are shown.

We implement the continuous phase-only synthetic discriminant function using the commercial liquid crystal television. The filter is a nonlinear combination of the component phase-only filters of the training images and is generated with the iterative method. It produces narrow correlation peaks at the origin without sidelobe problem. Experimental results are shown.

In this paper, we introduce a unified model-based pattern recognition approach that can be formulated into a variety of techniques to be used for a variety of applications. Complex phasor addition and cancellation are incorporated into the design of filter(s) to perform implicit logical operations using linear correlation operators. These implicit logical operations are suitable to implement high-level gray-scale morphological transformations of input images. In this way we effectively project nonlinear decision boundaries into the input signal space yet maintain the mathematical simplicity of linear filter designs. We apply this approach to the automatic distortion- and intensity-invariance object recognition problem. We introduce a set of shape operators or complex filters that are logically structured into a filter bank architecture to accomplish the distortion and intensity-invariant system. This synthesized complex filter bank is optimally sensitive to fractal noise representing natural scenery. The sensitivity is optimized for a specific fractal parameter range using the Fisher discriminant. The output responses of the proposed system are shown for target, clutter, and pseudo-target inputs to represent its discrimination and generalization capability in the presence of distortion and intensity variations.

Liquid crystal televisions have become increasingly popular as low-cost spatial light modulators. While the early devices suffered from poor resolution and low contrast, recent models compare favorably to the more traditional (and expensive) modulators. One of the most recent LCTVs is found in the InFocus TVT-6000 television projector. The panels in this projector have 480 X 440 pixels with a 1.32' diagonal clear aperture. A wavefront splitting interferometer has been constructed and analyzed for measuring the complex characteristics of these modulators, including phase and amplitude coupling. The results of this evaluation will be presented.

We propose an idea for security verification of credit cards, passports, and other ID so that they cannot easily be reproduced. A new scheme of complex phase/amplitude patterns that cannot be seen and cannot be copied by an intensity sensitive detector such as a CCD camera is used. The basic idea is to permanently and irretrievably bond a phase mask to a primary identification amplitude pattern such as a fingerprint, a picture of a face, or a signature. Computer simulation results and tests of the proposed system will be provided to verify that both the phase mask and the primary pattern are separately readable and identifiable in an optical processor or correlator.

Terrain-discrimination is an important component of wide-area surveillance, with applications to battlefield terrain and agricultural terrain. Recently a statistical method was proposed by Mahalanobis and Singh to design spatial filters to recognize and discriminate between various textures. We apply this technique to texture discrimination in SAR images. It is demonstrated that the technique is a viable one and its robustness for terrain discrimination in SAR imagery is shown. Spatial correlation filters are used for texture distinction. The filters are implementable as optical (or digital) correlators for fast real-time texture recognition without segmentation. The filter coefficients are determined via eigenvector analysis. Examples for airborne SAR images are given to illustrate the proposed scheme for terrain discrimination.

This paper presents an automatic in-line optical defect inspection systems for aperture grilles used for Sony Trinitron related products. The requirement to detect defects as small as 50 micrometers X 150 micrometers in as large as 1000 mm X 800 mm aperture grilles imposes a severe limitation for conventional machine vision techniques.

This paper reports results of a recent field experiment using a prototype system to evaluate the acousto-optic tunable filter polarimetric hyperspectral imaging technology for target detection applications.

One of the main factors determining the performance of a binary joint transform correlator is the method used to set the binarization threshold in the transform plane. An adaptive threshold can be set by computing the average of a small local window around each pixel. It is shown theoretically and verified experimentally that the optimum size of this window for a practical BJTC configuration is a single column wide and 3 - 5 pixels long. The detection peak amplitude in experimental runs using a local window was 2 - 4 times as large as in runs using the median of the Fourier plane as the threshold. Since these adaptive thresholds are easier to compute than the median of the entire Fourier plane, this technique can be used to improve BJTC performance while reducing system complexity.

An integrated automatic target recognition processor utilizing a feature extraction wavelet preprocessor and a pattern recognition neuroprocessor is introduced. Feature extraction of orientation and scale information using a Morlet wavelet is introduced. This feature extraction wavelet preprocessor is used for pre-attentive vision such that the orientation, scale, and location feature are effectively extracted. This wavelet preprocessor will enable the neuroprocessor to dynamically focus its vision on incoming targets based on their known features. This will result in higher discrimination and lower false alarm rate. The theoretical analysis of the orientation and scale selectivity using a Morlet wavelet is provided. Experimental demonstrations of wavelet preprocessing for feature extraction are also provided.

In this paper, we present an OptoNeural approach to the problem of multitarget tracking. The proposed hybrid OptoNeural system uses a new optical binary phase extraction joint transform correlator (BPEJTC) to reduce the massive input target data into a few correlation peak signals. A parallel computational neural network is then used for effective target tracking data association based on these correlation signals. For real-time operation, the BPEJTC is optically implemented using the high resolution LCD spatial light modulators and CCD detectors. Some experimental results on simultaneous tracking of multitargets are also provided.

A unipolar shift-invariant associative Hamming net is described in this paper. The proposed Hamming net is a three-layer neural network, in which the first layer is a shift-invariant Hamming layer using unipolar interconnection weight matrix, the second layer is a winner- take-all layer, and the last layer is a memory-mapping layer. In experiment, a hybrid optical architecture using photorefractive holograms is proposed.

We designed, fabricated, and tested an optically addressed spatial light modulator (SLM) that performs a 3 X 3 kernel image convolution using ferroelectric liquid crystal on VLSI technology. The chip contains a 16 X 16 array of current-mirror-based convolvers with a fixed kernel for finding edges. The pixels are located on 75 micron centers, and the modulators are 20 microns on a side. The array successfully enhanced edges in illumination patterns. We developed a high-level simulation tool (CON) for analyzing the performance of convolving SLM designs. CON has a graphical interface and simulates SLM functions using SPICE-like device models. The user specifies the pixel function along with the device parameters and nonuniformities. We discovered through analysis, simulation and experiment that the operation of current-mirror-based convolver pixels is degraded at low light levels by the variation of transistor threshold voltages inherent to CMOS chips. To function acceptable, the test SLM required the input image to have an minimum irradiance of 10 (mu) W/cm². The minimum required irradiance can be further reduced by adding a photodarlington near the photodetector or by increasing the size of the transistors used to calculate the convolution.

The Federal Bureau of Investigation (FBI) has recently sanctioned a wavelet fingerprint image compression algorithm developed for reducing storage requirements of digitized fingerprints. This research implements an optical wavelet transform of a fingerprint image, as the first step in an optical fingerprint identification process. Wavelet filters are created from computer- generated holograms of biorthogonal wavelets, the same wavelets implemented in the FBI algorithm. Using a detour phase holographic technique, a complex binary filter mask is created with both symmetry and linear phase. The wavelet transform is implemented with continuous shift using an optical correlation between binarized fingerprints written on a Magneto-Optic Spatial Light Modulator and the biorthogonal wavelet filters. A telescopic lens combination scales the transformed fingerprint onto the filters, providing a means of adjusting the biorthogonal wavelet filter dilation continuously. The wavelet transformed fingerprint is then applied to an optical fingerprint identification process. Comparison between normal fingerprints and wavelet transformed fingerprints shows improvement in the optical identification process, in terms of rotational invariance.

In spite of many advances of IR imaging technology that have been achieved, the detection of dim point targets from IR clutter backgrounds still remains a key problem in real-time IR systems. We present a new detection scheme based on a linear detector and an improved probabilistic neural network classifier for small SNR, moving point targets detection in strong IR noise and clutter backgrounds. Computer simulation was conducted and simulation results confirmed the validity of the detection scheme.

Fourier processing method is very popular in signal and image analysis. In this paper, a generalized Fourier transform (GFT) processor is introduced. It is generalized in that it contains the conventional Fourier transform (CFT) processor as a very limited special case by extending the diagonal filter matrix in CFT to a nondiagonal one. Thus, obviously, the GFT processor should be more powerful than the CFT processor. A computer demonstration shows this.

This paper reviews the basic principles, physical processes, and most recent demonstrations of optical correlators using dynamic holographic techniques in nonlinear media such as bulk photorefractives (PR), thin multiple quantum wells (MQW), and bacteriorhodopsin (BR) films.

A photorefractive mutually pumped phase conjugator is operated in both the steady-state and transient regimes to demonstrate several important optical processing applications. While in the transient regime, or when photorefractive grating formation is taking place in the mutually pumped conjugator, cross talk is observed between the two input beams. This feature is used to demonstrate the instantaneous transfer of pictorial information from one laser beam to another and to perform the correlation of moving objects. Image transfer, optical correlation, and interferometric applications are presented. Key issues concerning the spatial resolution, time response, and Bragg condition are discussed for these applications.

In this paper, application of thick photorefractive (PR) crystal to wavelength-multiplexed reflection-type matched filter is described. The Bragg diffraction limitation and the shift- invariant property using thick volume PR crystal are discussed, in which we see that the Bragg diffraction is severely limited by the thickness of the crystal. To alleviate the Bragg limitation in terms of shift tolerance, we have shown that the reflection-type wavelength-multiplexed PR filter can be used. Experimental confirmation of these findings are also provided.

Full optical implementation of band-pass, band-reject and high-pass temporal frequency spatial filters for amplitude images are proposed. These filters rely on the exponential temporal response of photorefractive crystals and the interference with phase conjugate beams. Band- pass and high-pass filters may be realized using a phase conjugate Michelson interferometer which has a difference in the response times between two arms, and band-reject filter using the interference between the output from a band-pass filter and a reference beam. The frequency characteristics of these filters are analyzed with numerical examples, and an analogy of these optical filters to corresponding electronic circuits is introduced.

In this paper we present a class of nonlinear wavelet transforms (WT) that are invariant under scale, rotation, and shift transformation of the input image. We provide an analytical proof of the invariance property of the WT as well as numerical simulation supporting this concept.

Novel optical wavelet correlators are presented in this paper. Instead of computing the correlation between the target and the reference patterns, these optical wavelet correlators compute the correlation between the wavelet transform coefficients of the target and reference patterns. It incorporates the efficient feature detection and noise suppression capabilities of wavelet transformation and the inherent high sped and 2D nature of objects. The feature extraction and target identification operations are performed in a single step. By suitably choosing the wavelet functions and their dilation factors, these wavelet correlators are capable of effectively recognizing targets in the presence of clutter and noise. Architectures for both Vander Lugt-type wavelet correlator and wavelet joint transform correlator are discussed in the paper. A wavelet function especially suitable for optical implementation is described. Preliminary experimental results are provided.

A desktop design and manufacturing system for binary diffractive elements, MacBEEP, was developed with the optical researcher in mind. Optical processing systems for specialized tasks such as cellular automation computation and fractal measurement were constructed. A new family of switchable holograms has enabled several applications for control of laser beams in optical memories. New spatial light modulators and optical logic elements have been demonstrated based on a more manufacturable semiconductor technology. Novel synthetic and polymeric nonlinear materials for optical storage are under development in an integrated memory architecture. SBIR programs enable creative contributions from smaller companies, both product oriented and technology oriented, and support advances that might not otherwise be developed.

3D silicon technology has been under development since 1980, primarily aimed at on-focal- plane signal processing to solve a variety of military sensor systems problems. The thrust has been to bring more and more parallel analog and digital processing into the closest possible proximity to the detector array. At this time on-focal-plane functionality includes preamplification, spatial and temporal matched filtering, nonuniformity correction, neural networks, analog-digital conversion, digital logic, and digital memory. Historically, a custom- built specialty technology constrained by cost in its applicability, 3D silicon has undergone a dual-use conversion to include high-volume, low-cost commercial computer electronics. 3D silicon is on the way to becoming the lowest-cost-per-gate technology available and, because of this, sensor system design and performance will be revolutionized.

We have made 128 X 128 and 256 X 256 spatial light modulators using active backplanes fabricated through a commodity silicon foundry and incorporating a thin ferroelectric liquid crystal light modulating layer at the backplane's surface by means of postprocessing of individual foundry die. These electrically addressed devices exhibit optical rise and fall times as short as 105 microsecond(s) , with contrast ratios in images as high as 100:1, and in zero-order diffracted light as high as 200:1. Total optical throughput to the zero-order diffracted beam exceeds 10% for the 256 X 256 devices and 17% for the 128 X 128 devices. Frame update times shorter than 100 microsecond(s) , corresponding to image information throughput of greater than 80 MBytes/s, were realized by employing pipelining techniques in conjunction with a wide digital input word.

Foster-Miller is developing a new family of SLM devices, based on the Franz-Keldysh (FK) effect in bulk III-V semiconductors, for several applications in optical signal processing and switching. Spatial light modulators constructed using the FK effect offer contrast ratios and switching energies rivalling state-of-the-art devices. However, our devices require typically only 2 or 3 bulk epilayers (available commercially), significantly reducing the materials and fabrication cost. Using the FK effect in an asymmetric Fabry-Perot geometry, we demonstrate high contrast (120:1 best, 50:1 typical) with drive voltages roughly 10 - 20 V and 2 - 3 nm optical bandwidth. We demonstrate high-contrast reflection mode optical modulation at a number of wavelengths in the 800 - 950 nm band.

An optical angular mapping for invariant optical/digital pattern recognition is presented using self-pumped phase conjugation followed by a vertical projection. In this method, the vertical contour features of an object are most enhanced and vertically projected at each angle as the object is rotated. The vertical enhancement of the contour and the projection are optically performed with self-pumped phase conjugation and a cylindrical lens, respectively. The peaks in the resulting projection at the angle are extracted and summed to produce an angular signature function. The signature function is then energy-normalized for scale invariance and pattern recognition is performed by calculating a crosscorrelation between two signature functions. The method exhibits invariance to shift, rotation, and variations in scale.

Commercially available rewritable magneto-optic compact disks are utilized as input spatial light modulators in a hybrid optical/digital joint transform correlator and in a purely optical correlator. Identical images are written on two disks using a commercially available optical disk drive, and good auto-correlation signals were obtained.