Mapping the mesoscale interface structure in polycrystalline materials.

A new experimental approach to the quantitative characterization of polycrystalline microstructure by scanning electron microscopy is described. Combining automated electron backscattering diffraction with conventional scanning contrast imaging and with calibrated serial sectioning, the new method (mesoscale interface mapping system) recovers precision estimates of the 3D idealized aggregate function G(x). This function embodies a description of lattice phase and orientation (limiting resolution approximately 1 degree) at each point x (limiting spatial resolution approximately 100 nm), and, therefore, contains a complete mesoscale description of the interfacial network. The principal challenges of the method, achieving precise spatial registry between adjacent images and adequate distortion correction, are described. A description algorithm for control of the various components of the system is also provided.

A closed-form neural network for discriminatory feature extraction from high-dimensional data.

We consider a new neural network for data discrimination in pattern recognition applications. We refer to this as a maximum discriminating feature (MDF) neural network. Its weights are obtained in closed-form, thereby overcoming problems associated with other nonlinear neural networks. It uses neuron activation functions that are dynamically chosen based on the application. It is theoretically shown to provide nonlinear transforms of the input data that are more general than those provided by other nonlinear multilayer perceptron neural network and support-vector machine techniques for cases involving high-dimensional (image) inputs where training data are limited and the classes are not linearly separable. We experimentally verify this on synthetic examples.

Modulation transfer function measurement method for electrically addressed spatial light modulators.

The modulation transfer function (MTF), when used with amplitude modulation (m(A)) data, is a vital coherent optical performance measure for a spatial light modulator (SLM). A new image plane amplitude MTF (MTF(A)) measurement method is presented for electrically addressed SLMs. It involves digital analysis of the output image of a square-wave pattern written onto the SLM. Modulation-level effects are also addressed. Optical laboratory results are presented for two liquid-crystal SLMs. The need to consider amplitude rather than intensity modulation (when coherent optical processing applications are considered) is noted in terms of SLM biasing.

Quadratic Gabor filters for object detection.

We present a new class of quadratic filters that are capable of creating spherical, elliptical, hyperbolic and linear decision surfaces which result in better detection and classification capabilities than the linear decision surfaces obtained from correlation filters. Each filter comprises of a number of separately designed linear basis filters. These filters are linearly combined into several macro filters; the output from these macro filters are passed through a magnitude square operation and are then linearly combined using real weights to achieve the quadratic decision surface. For detection, the creation of macro filters (linear combinations of multiple single filters) allows for a substantial computational saving by reducing the number of correlation operations required. In this work, we consider the use of Gabor basis filters; the Gabor filter parameters are separately optimized. The fusion parameters to combine the Gabor filter outputs are optimized using an extended piecewise quadratic neural network (E-PQNN). We demonstrate methods for selecting the number of macro Gabor filters, the filter parameters and the linear and nonlinear combination coefficients. We present preliminary results obtained for an infrared (IR) vehicle detection problem.

Detection filters and algorithm fusion for ATR.

Detection involves locating all candidate regions of interest (objects) in a scene independent of the object class with object distortions and contrast differences, etc., present. It is one of the most formidable problems in automatic target recognition, since it involves analysis of every local scene region. We consider new detection algorithms and the fusion of their outputs to reduce the probability of false alarm P(FA) while maintaining high probability of detection P(D). Emphasis is given to detecting obscured targets in infrared imagery.

Nonlinear optical hit-miss transform for detection.

Morphological processing involves nonlinear low-level image-processing operations that can be realized on optical processors. Amodified version of the hit-miss morphological transform is described for object detection. Simulation results and optical laboratory realizations are presented. Some of the simple filters required can be realized as ternary-phase-amplitude optical filters.

Symbolic substitution modified signed-digit optical adder.

A high-accuracy fixed-point optical adder that operates in parallel on many long words and that uses a pipelined correlator architecture is described. A symbolic substitution algorithm with the modified signed-digit number representation is used to perform fixed-point additions with limited carries. A new set of substitution rules and encodings is developed to combine the recognition and substitution steps into one correlation operation. This reduces hardware requirements, improves throughput by reducing the space-bandwidth product needed, and reduces latency (the delay between when data enter the processor and when the final output is available) by a factor of 2. This algorithm and our new modified signed-digit encodings and substitution rules improve the performance of other correlator and noncorrelator optical numeric computing architectures.

High-accuracy pipelined iterative-tree optical multiplication.

A high-accuracy optical multiplier that uses an optical correlator is described. A symbolic substitution adder that uses the modified signed-digit number representation is used as the basic module. Emphasis is placed on the multiplication of many long words in parallel with minimum latency. The encoding method we employ in the adders permits the use of a new optical algorithm and architecture to generate partial products in symbolic form in parallel. Our multiplication algorithm and architecture are shown to be preferable to other optical techniques and to be competitive with digital technology; they are also shown to be particularly attractive for matrix-vector multiplication applications.

One-dimensional collimation of laser array outputs.

We consider a computer-generated hologram for the one-dimensional collimation in x of the output from a linear laser-diode array in y. Our concern is to produce one-dimensional pencil beams from each laser diode with small cross talk between the output from the separate laser diodes. Such outputs can be used in matrix-vector, neural net, and interconnection applications. The efficiency and the design of the computer-generated hologram are detailed, and initial optical laboratory results with an electron-beam recorded computer-generated hologram are presented.

Multitarget-tracking optical processing system.

We consider multitarget tracking, estimating the state vector (two-dimensional position and velocity) of each target from physical measurements. We consider a full system for this and the role for analog optical processing within its subsystems. We emphasize the neural network data-association subsystem (which associates measurements in the present input frame with estimates from previous frames of data). Our new optimization neural net results concern associations between measurements and estimates and show that use of a simple fixed-coefficient estimation filter is sufficient. For completeness in our full system approach we briefly describe our optical detection subsystem and its use to reduce frame-to-frame jitter in the measurements. We also briefly note our Hough-transform optical subsystem and discuss its use in detecting and correcting data dropout errors and errors in the data-association and estimator systems. We conclude that analog optical processing has significant use in a full multitarget tracking system.

High-capacity neural networks on nonideal hardware.

We present a new training-out algorithm for neural networks that permits good performance on nonideal hardware with limited analog neuron and weight accuracy. Optical neural networks are emphasized with the error sources including nonuniform beam illumination and nonlinear device characteristics. We compensate for processor nonidealities during gated learning (off-line training); thus our algorithm does not require real-time neural networks with adaptive weights. This permits use of high-accuracy nonadaptive weights and reduced hardware complexity. The specific neural network we consider is the Ho-Kashyap associative processor because it provides the largest storage capacity. Simulation results and optical laboratory data are provided. The storage measure we use is the ratio M/N of the number of vectors stored (M) to the dimensionality of the vectors stored (N). We show a storage capacity of M/N = 1.5 on our optical laboratory system with excellent recall accuracy, > 95%. The theoretical maximum storage is M/N = 2 (as N approaches infinity), and thus the storage and performance we demonstrate are impressive considering the processor nonidealities we present. Our techniques can be applied to other neural network algorithms and other nonideal processing hardware.

Morphological and wavelet transforms for object detection and image processing.

We consider the problem of detecting multiple distorted objects in an input scene with clutter. The input scenes contain different types of background clutter and multiple objects in different classes, with different object aspect views, different object representations, hot/cold/bimodal/partial object variations, and high/low contrast object variations. Several new optical morphological operations for use in the above detection problem and in other general low-level image-processing applications are described, and several examples of their use are provided. For difficult detection problems in which high detection rates and low false-alarm rates are required we combine morphological operations and optical wavelet transforms to reduce clutter and improve object detection. The details of this set of filters and initial testresults are given. The most computationally demanding operations required in all cases are realizable on an optical correlator.

High-speed acousto-optic mapping modulator for the generalized Hough transform.

We describe a high-speed acousto-optic Hough-transform mapping modulator. This mapping modulator generates any θ slice of the straight line Hough-transform and one-dimensional slices of generalized Hough transforms (e.g., for circles and ellipses). We derive the mapping functions for the acousto-optic modulator for both circle and ellipse Hough transforms and show simulations of generalized Hough transforms using both functions. We also describe how the mapping modulators can compute Hough transforms for nonanalytically describable inputs.

Optical hit-miss morphological transform.

We present an optical correlator implementation of the morphological hit-miss transform. This provides improved recognition of objects in clutter compared with standard optical pattern-recognition correlator techniques. The hit-miss transform is modified to use rank-order filtering since this gives better performance in noise and clutter. By varying the correlation plane threshold, we can perform dilations, rank-order filters, and erosions on the same multifunctional coherent optical correlator system. We quantify the thresholds required for generic object part recognition and provide simulated and optical laboratory data.

Multitarget data association using an optical neural network.

A neural network solution to the data association problem in multitarget tracking is presented. It uses position and velocity measurements of the targets over two consecutive time frames. A quadratic neural energy function, which is suitable for an optical processing implementation, results. Simulation resultsusing realistic target trajectories with target measurement noise including platform movement or jitter are presented. The results show that the network performs well when track data are corrupted by significant noise. Several possible optical neural network architectures to implement this algorithm are discussed, including a new all-optical matrix-vector multiplication approach. The matrix structure is employed to allow binary-ternary spatial light modulators to be used.

Optical correlator production system neural net.

A new neural net is described that can easily and cost-effectively accommodate multiple objects in the field of view in parallel. The use of a correlator achieves shift invariance and accommodates multiple objects in parallel. Distortion-invariant filters provide aspect-invariant distortion. Symbolic encoding, the use of generic object parts, and a production system neural net allow large class problems to be addressed. Optical laboratory data on the production system inputs are provided and emphasized. Test data assume binary inputs, although analog (probability) input neurons are possible.

Advanced distortion-invariant minimum average correlation energy (MACE) filters.

The original minimum average correlation energy (MACE) filter is addressed by using a new database (strategic relocatable objects, missile launchers) and including noise performance, depression angle, and resolution effects on the number of training set images that are required. Major attention is given to our new MACE filter algorithms for distortion-invariant pattern recognition: shifted-MACE filters (to suppress large false correlation peaks), minimum variance-MACE filters (for improved noise performance), multiple symbolic encoded filters (to reduce the effect of false correlation peaks), and Gaussian-MACE filters (to improve noise performance and intraclass recognition and reduce the training set size).

Minimum noise and correlation energy optical correlation filter.

A new distortion-invariant optical correlation filter to produce easily detectable correlation peaks in the presence of noise and clutter and to provide better intraclass recognition is presented. The basic ideas of the minimum variance synthetic discriminant function correlation filter (which minimizes noise variance in the output correlation peak/plane) and the minimum average correlation energy filter (which minimizes the average correlation plane energy over all the training images) are unified in a new filter that produces sharp correlation peaks while maintaining an acceptable signal-to-noise ratio in the correlation plane output. This new minimum noise and correlation energy filter approach introduces the concept of using the spectral envelope of the training images and the noise power spectrum to obtain a tight bound to the energy minimization problem that is associated with distortion-invariant filters in noise while allowing the user a variable parameter to adjust depending on the noise or clutter that is expected. We present the mathematical basis for the minimum noise and correlation energy filter and the initial simulation results.

Invariance and neural nets.

Application of neural nets to invariant pattern recognition is considered. The authors study various techniques for obtaining this invariance with neural net classifiers and identify the invariant-feature technique as the most suitable for current neural classifiers. A novel formulation of invariance in terms of constraints on the feature values leads to a general method for transforming any given feature space so that it becomes invariant to specified transformations. A case study using range imagery is used to exemplify these ideas, and good performance is obtained.

Performance evaluation of minimum average correlation energy filters.

A new SDF type correlation filter referred to as the minimum average correlation energy (MACE) filter has been recently described in the literature. In this paper, we experimentally address the distortion tolerance and noise properties of this filter. The MACE filter has attractive properties that include: easily detectable peaks, distortion invariance, simplified training set selection, solutions to input bias effects, performance in noise and real background clutter, and less clutter with its reduced number of training set images. Each of these properties is investigated in detail in this paper.