In this paper we design an optimum receiver to detect a pattern or a reference signal. We design a receiver which detects the signal distorted by a multiplicative noise on the signal itself, as well as by additive noise and by non- overlapping scene noise. We use hypothesis testing to determine the location of a target buried in multiplicative, additive, and non-overlapping background noise. We also consider the case when the statistics of the multiplicative noise are not available. When additive noise is present, the usual maximum likelihood estimation method gives rise to nonlinear system of equations. We present a vectorial data fitting method to estimate the unknown parameters. Using computer simulations we show that, for the images tested here, the optimum receiver performs better than some of the existing receivers.

Two-dimensional correlation between a reference template and an input scene is a powerful pattern recognition technique but is demanding of computational power. Coherent optical correlators, exploiting the Fourier transforming properties of a lens and the capability to impart a phase modulation on a wavefront with an appropriate spatial light modulator (SLM), hold the promise of real-time implementation of two- dimensional correlation for realistic pattern recognition problems. However, their practical use has been delayed in many applications by the lack of availability of suitable SLM devices with the required speed and dynamic range, with different needs for input and frequency plane modulators. It is now possible to compute a two-dimensional Fourier transform at video-rates with various digital signal processing chip sets. Thus a hybrid correlator is proposed in which the input scene is digitally Fourier transformed at video-rate, and multiple templates searched during the next video frame interval by optical mixing and Fourier transformation at a speed at least two orders of magnitude faster than possible with digital methods. In this way, the input SLM is avoided and a precise spectrum is available for subsequent digital or optical mixing with the stored templates. The speed advantage over all-digital processing allows unconstrained pattern recognition problems to be tackled that require many template searches to match the input with a reference function. Different hybrid correlator configurations are considered, together with discussion of the various digital chip sets available to perform the video-rate FFT, as well as the SLM devices currently available that are suitable as frequency domain phase modulators.

Classification of real-time X-ray images of randomly oriented touching pistachio nuts is discussed. The ultimate objective is the development of a system for automated non-invasive detection of defective product items on a conveyor belt. We discuss the extraction of new features that allow better discrimination between damaged and clean items (pistachio nuts). This feature extraction and classification stage is the new aspect of this paper; our new maximum representation and discriminating feature (MRDF) extraction method computes nonlinear features that are used as inputs to a new modified k nearest neighbor classifier. In this work, the MRDF is applied to standard features (rather than iconic data). The MRDF is robust to various probability distributions of the input class and is shown to provide good classification and new ROC (receiver operating characteristic) data.

We propose and demonstrate an all-optical pulse generator which is based on regenerative pulsation in optical bistable systems. The novel device consists of a nonlinear interference filter, a photorefractive crystal and an optical feedback loop. In such a pulse generator, when the input beam intensity exceeds the switch-on threshold of the system, a train of square-wave pulses will be produced. The firing rate is an increasing function of the intensity of the input beam. The principle of operation and the experimental results are presented and discussed.

This paper examines the use of composite filters for improving the effectiveness of a Vanderlugt correlator when used for fingerprint identification. A digital simulation, which accounts for noise sources in the optical setup, is used to design and test composite matched spatial filters. Results are presented for a real time video image database containing 10 seconds of video from 200 fingers. Using the composite matched spatial filter the Vanderlugt correlator is getting 70% correct identifications with no false positives.

We report experimental laboratory results using filters that optimize the Rayleigh quotient [Richard D. Juday, 'Generalized Rayleigh quotient approach to filter optimization,' JOSA-A 15(4), 777-790 (April 1998)] for discriminating between two similar objects. That quotient is the ratio of the correlation responses to two differing objects. In distinction from previous optical processing methods it includes the phase of both objects -- not the phase of only the 'accept' object -- in the computation of the filter. In distinction from digital methods it is explicitly constrained to optically realizable filter values throughout the optimization process.

In this paper we compare the performance of linear and nonlinear filters on binary images. The maximum-likelihood ratio test (MLRT) processor is optimal for detection. We generalize it and use it for location. It is thus well suited as an upper reference with which we compare the performance of the other filters. We present the maximum likelihood ratio approximation (MLRA) and we compare the MLRA and the nonlinear joint transform correlator (NLJTC) with the MLRT. By having a look at the impulse responses of these processors we can explain the similarities in performance. We also compare the MLRT with classical linear filters, such as the optimal trade- off (OT) filter and the classical matched filter (CMF). We show that the automatic regularization given by the binary noise makes the CMF perform almost as good as the MLRT. The OT filter can be regularized through the choice of the tuning parameter and it also shows almost as good performance as the MLRT.

In this paper, we incorporate the multivariate statistical methods into an incoherent optical correlator based optoelectronic pattern recognition system and realize the distortion-invariant recognition. In this approach, a set of eigenimages are first extracted from a large number of training images including various typical distortions by using the principal component analysis and then are used as the reference patterns in the correlator. The optical correlation results between the testing image and the set of eigenimages construct a feature space, on which the multivariate discriminant analysis is performed. During both the training and the classification process, a bifurcating tree structure is used, by which the recognition speed of the system can be greatly improved.

Recent and near-term advancements in our multi-level (analog) phase/amplitude liquid crystal spatial light modulators will be presented. These advancements include higher resolution, smaller pixel pitch, planarized pixel pads, and higher speed modulation for phase-only, amplitude-only, and phase- amplitude-coupled modulation. These devices have applications in optical processing, optical storage, holographic display, and beam steering. Design criteria and experimental data will be presented.

We review GaAs Fabry-Perot vertical cavity multiple quantum well (MQW) spatial light modulators (SLMs) developed at Sanders, a Lockheed Martin Company, and demonstrate their use in optical image correlation. These MQW SLMs are reflective- mode modulators using electrically-tunable absorption to modulate the reflected intensity. The operation of the MQW SLMs with a newly-developed Labview graphical user interface is described. A compact Vander-Lugt image correlator is described which was configured using MQW SLMs: binary 128 X 128 pixel image input with a binary filter plane. In addition, the architecture of 8-bit gray-scale MQW-SLM module developed at Sanders is discussed. The performance of the image correlator was characterized using amplitude-encoded binary phase-only filters and various test targets including test imagery supplied by US Army AMCOM, and is compared with simulations for peak-to-secondary efficiencies on these data. Finally, high-speed (250,000 frames per second) target recognition of 128 X 128 pixels binary input imagery is demonstrated.

Some methods of quantifying the utility of a spatial light modulator for optical correlation pattern recognition are introduced and compared with earlier methods. An SLM should be judged according to the problem it is being used in, and the metric optimized in the filter construction can be used to measure the utility of the SLM.

We investigate the possibility of using a real-axis spatial light modulator (SLM) to realize complex-amplitude modulation with full coverage over the unit circle. The real-axis SLM produces a pixelated bipolar-amplitude wavefront. Each pixel is basically a spatial pulse width with the signal information being carried by the pulse's amplitude. Data streams generated in this manner have real and imaginary components due to the relative even and odd symmetry of the pulse amplitude modulation. When the pulse rate is twice the minimum Nyquist rate for band-limited amplitude modulation, it is possible to resolve the signal into its quadrature phase components (real and imaginary terms). By changing the relative amplitude of these quadrature phase terms, any value within the complex plane is accessible. Since the SLM does not have any optical gain, coverage is limited to the unit circle in the complex plane.

The measurement of two important performance limitations of commonly available spatial light modulations is addressed. A method for the determination of modulation transfer function and amplitude modulation from CCD captures of imaged gratings is presented. An interferometric technique is employed for estimating spatial non-uniformities and optical non-flatness of a liquid crystal panel.

Recent developments have shown that it is possible to create an optical correlator, or recognizer, based on ferroelectric liquid crystal spatial light modulator technology. In this paper we present the expansion of this idea to include an in- house FLC over silicon SLM that allows a compact recognizer to be built. The basic architecture is the binary 1/f correlator built around a high-speed 320 X 240 pixel FLC over silicon SLM. Such an architecture can be built with low tolerance optomechanics to make a powerful yet robust optical recognizer. Initial result from this recognizer are presented in this paper along with a discussion of how this architecture can be used in more abstract recognition tasks such as motion estimation instead of the more traditional 'tank-in-the-bush' scenario.

Spatial light modulators used as inputs to optical correlators often modulate both the amplitude and phase of the coherent illumination. This means that matched filters commonly used are not truly matched to the scenes displayed on these modulators. Filters are normally calculated from an assumed amplitude-only input. A few analytically manageable examples are given which demonstrate the errors that can occur and suggestions are made for using the phase modulating characteristics to an advantage.

A new scheme for parallel optical decryption and the display of encrypted image information is presented. The scheme is based on a common path interferometer configuration providing a simple and robust optical setup. Images are encrypted directly during recording by use of a combined phase encoding and phase scrambling method. The encoding and encryption does not require sophisticated, iterative and time consuming optimization algorithms. Pixels are independently encoded and encrypted by use of a simple look-up table technique that can be performed in milliseconds on a standard personal computer. Optical decryption can subsequently be implemented in the common path interferometer by use of single phase or if desired a combined phase/amplitude key. An advantage of the presented method is that the encrypted image may selectively require recording and decryption of phase values or amplitude values or a combination thereof. Another advantage is that decryption is performed in a plane adjacent to the encrypted mask or an equivalent plane whereby generation of speckles in the decrypted image is strongly suppressed. Finally, there is no requirement of positioning a decrypting mask or spatial light modulator in the optical Fourier plane whereby an accurate three dimensional positioning requirement can be avoided.

The combination of a high resolution charge-coupled device camera and a phase modulating liquid crystal array allows real-time holographic (interference) fringes to be detected and displayed. The original object may then be reconstructed optically. A modified Mach-Zehnder (Leith-Upatnieks) architecture is utilized for producing the required small object-to-reference beam angle. The hologram is updated and reconstructed at television field rates. Several analog reconstructions are presented as examples of the performance of the simple system. If the input scene is optically Fourier transformed, the system may be used to produce real-time Fourier transform matched filters. These filters can be displayed immediately in an optical correlator for pattern recognition applications.

A new computational paradigm is introduced. Other image representation such as Fourier transforms and wavelet decompositions depend on linear superposition of basis functions. The k-factor image factorization reduces an image into a finite or infinite set of contrast-ordered images whose joint product reproduces the original image. It is experimentally found that shadows and noise often fall into factors disjoint from the 'pure' image. The analytical foundations of the k-factor method are given, followed by full factorizations and reconstructions, and future research directions are described that include shadow removal, speckle reduction, medical and military image analysis, and commercial applications.

We present an optical feature extraction system with a microlens array processor. The system is suitable for online implementation of a variety of transforms such as the Walsh transform and DCT. Operating with incoherent light, our processor accepts direct optical input. Employing a sandwich- like architecture, we obtain a very compact design of the optical system. The key elements of the microlens array processor are a square array of 15 X 15 spherical microlenses on acrylic substrate and a spatial light modulator as transmissive mask. The light distribution behind the mask is imaged onto the pixels of a customized a-Si image sensor with adjustable gain. We obtain one output sample for each microlens image and its corresponding weight mask area as summation of the transmitted intensity within one sensor pixel. The resulting architecture is very compact and robust like a conventional camera lens while incorporating a high degree of parallelism. We successfully demonstrate a Walsh transform into the spatial frequency domain as well as the implementation of a discrete cosine transform with digitized gray values. We provide results showing the transformation performance for both synthetic image patterns and images of natural texture samples. The extracted frequency features are suitable for neural classification of the input image. Other transforms and correlations can be implemented in real-time allowing adaptive optical signal processing.

The joint transform correlator (JTC) is a robust architecture for optical object recognition or verification and has been investigated intensively for a number of years. In order to improve its performance digital processing, such as binarization to an appropriate threshold, has been proposed at the recording stage of the joint power spectrum (JPS), this of course to some extent negates the benefits of an optical system. We propose to employ a third beam which is coherent to the beams forming the JPS and is the Fourier transform of an appropriate third input image and thus can be used to modify the JPS so as to implement a phase only filtering action for example. The triple joint transform correlator thus consists of three Fourier spectra incident simultaneously on a photorefractive crystal and can exhibit improved discrimination over the conventional JTC without an intermediate digital processing stage. Computer simulations and experimental results are presented.

Problem of optical pattern recognition (OPR) is sorting numerous images into several subsets. A certain generalized image is ascribed to each subject. For example, the letter 'A,' written in different fonts, to some generalized image of this letter. In this process one has to deprive it of individual attributes i.e. to narrow down its spatial spectrum. This result in a generalized object's image separation with informative fragments (IF) at places with sharp form's changing typical to many similar images. It is shown that in Nature such process can be realized by existing periodic defocusing of the crystalline lens, which leads to generalization of an image projected on the retina. Such process i.e. image defocusing can be used in technique for generalization real images in OPR systems. It was proposed the scheme of possible variant of such robot 'drawing' the generalized images of real objects. The results of computer imitation of 'drawing' generalized images by such a robot, are presented. In report also will be shown an example of possible solving the reverse task: expanding the spatial spectrum of the object with the poor own spectrum for alleviation the problem of its detection (for example, optical feature extractions for target identification). Such procedure according to Shenon's Theory cannot improve the real object's information, but permits alleviate detection of such object. An example of solving one such problem, is presented.

This paper introduces a system that provides estimations for the 2-D or 3-D position of a point target, with high spatial resolution. The system contains a conventional imaging lens being attached to a special diffractive optical element. The super resolved position estimation to be utilized with pixelated detector arrays is obtained due to the use of this optical element, which replicates the imaged point source on the detector array plane. A relatively simplified computation algorithm applied on the composite image yields the desired position estimation.

The area of information processing has grown dramatically over the last 50 years. In the areas of image processing and information storage the technology requirements have far outpaced the ability of the community to meet demands. The need for faster recognition algorithms and more efficient storage of large quantities of data has forced the user to accept less than lossless retrieval of that data for analysis. In addition to clutter which is not the object of interest in the data set, often the throughput requirements forces the user to accept 'noisy' data and to tolerate the clutter inherent in that data. It has been shown that some of this clutter, both the unavoidable clutter (clouds, trees, etc.) as well as the noise introduced on the data by processing requirements can be modeled as fractal or fractal-like. Traditional methods, using Fourier deconvolution on these sources of noise in frequency space, lead to loss of signal and can, in many cases, completely eliminate the target of interest. One parameter used to characterize fractal-like noise, the fractal dimension, has been investigated and fractal dimension images are presented.

An improved technique for multiple target detection using wavelet transform based joint transform correlation is proposed. In this technique, a bank of wavelet filters are generated using the Mexican hat wavelet function. These filters are then sequentially superimposed on the joint power spectrum before applying the inverse Fourier transform to yield the correlation output. This technique is found to yield high correlation discrimination for multiple target detection of gray level input scenes while avoiding the false alarms as well as rejecting the non-target objects. An all-optical implementation for the proposed wavelet JTC technique is also suggested. Computer simulation results confirm the effectiveness of the proposed technique.

As the interest in fractal geometry rises, the applications are getting more and more numerous in many domains. This paper deals with the problem of recognizing and classification to optical character recognition. For this purpose, we present a new method of feature extraction based on the principles of fractal geometry and wavelet. This allows us to establish a classification of Chinese character in order to apply to each of the isolated categories the most adapt recognition methods. In particular, the proposed method reduces the dimensionality of a two-dimensional pattern by way of a central projection approach, and thereafter, performs Daubechies' wavelet transformation on the derived one-dimensional pattern to generate a set of wavelet transformation sub-patterns, namely, curves that are non-self-intersecting. Further from the resulting non-self-intersecting curves, the divider dimensions are computed with modified box-counting approach. These divider dimensions constitute a new feature vector for the original two-dimensional pattern, defined over the curve's fractal dimensions. We have conducted several experiments in which a set of printed alphanumeric symbols and Chinese characters of varying fonts and orientation were classified, based on the formulation of our new feature vector. The results obtained from these experiments have consistently shown the character recognition method with the proposed feature vector can yield an excellent classification rate 100%.

One of the problems in the implementation of the Projection- Slice filter is the reduction of redundant information during filter synthesis. The wavelet transform is utilized to minimize the information redundancy. The method for this type of filter synthesis is discussed and simulation results of the algorithms involved are presented.

The new class of computer systems is presented, the architecture of which is controlled by parameters of the input images and based on usage of the adaptive moment image features (MIF) statistical analysis. Main types of MIF, as well as opportunity of their formation are described.

A multi-scale wavelet-based non-parametric algorithm for detecting and locating corners in 2D images is proposed. First using zero-crossing-based 2D edge detector we can get the edge elements, after edge linking and give each planar curve its orientation space representation we can get the orientation curves. Based on the multi-scale wavelet transform of the orientation curves we can utilize the information of local maximum positions to detect and locate the corners. Experimental results with some synthetic and real images show that this algorithm has high precision and stability of the corner detection and location.

A new algorithm based on scale-space median and gabor filtering is used to find boundaries in electron microscopy images under noise and low contrast. Boundary information from different scales are fused to find triple junctions and dihedral angles that are of use in material science. The proposed algorithm has been tested on various electron microscopy images and has been shown to be robust to changing imaging conditions, noise and contrast.

Today, the paper document is fast becoming a thing of the past. With the rapid development of fast, inexpensive computing and storage devices, many government and private organizations are archiving their documents in electronic form (e.g., personnel records, medical records, patents, etc.). In addition, many organizations are converting their paper archives to electronic images, which are stored in a computer database. Because of this, there is a need to efficiently organize this data into comprehensive and accessible information resources. The Imaged Document Optical Correlation and Conversion System (IDOCCS) provides a total solution to the problem of managing and retrieving textual and graphic information from imaged document archives. At the heart of IDOCCS, optical correlation technology provides the search and retrieval capability of document images. The IDOCCS can be used to rapidly search for key words or phrases within the imaged document archives and can even determine the types of languages contained within a document. In addition, IDOCCS can automatically compare an input document with the archived database to determine if it is a duplicate, thereby reducing the overall resources required to maintain and access the document database. Embedded graphics on imaged pages can also be exploited, e.g., imaged documents containing an agency's seal or logo, or documents with a particular individual's signature block, can be singled out. With this dual capability, IDOCCS outperforms systems that rely on optical character recognition as a basis for indexing and storing only the textual content of documents for later retrieval.

The general advantages of optical filtering methods for real- time image analysis have been extensively discussed in literature. Regarding the applicability of optical filtering for industrial purposes it is important that optical and opto- electronic off-the-shelf components can be used for the production of opto-electronic image processing hardware. However, practical experiences show that such components often do not realize theoretically designed filtering systems in an acceptable manner. Instead, due to their potential imperfection, they can be a source of noise effects and artifacts in filtered images. In optical inspection applications such artefacts can cause false alarms which cannot be tolerated. In this paper we focus on an opto- electronic image analysis system which has been composed using off-the-shelf opto-electronic components. We discuss the setup and its elements. The disturbing effects introduced by these components on the filtered images are demonstrated in a specific application example. Exploiting the fact that the noise effects of the optical components can be recorded, we implemented a digital forward-error-correction that reduces noise in the filtered images drastically. This method is very simple, fast, and efficient. Therefore, it is very useful in the development of opto-electronic image processing hardware for industrial purposes.

We propose an optical correlator system using volume hologram for database of matched filter. Optical correlator has high speed and parallel processing characteristics of optics. Matched filters are recorded into a volume hologram that can store data with high density, transfer them with high speed, and select a randomly chosen data element. The multiple reference images of database are prerecorded in a photorefractive crystal in the form of Fourier transform images, simply by passing the image displayed in a spatial light modulator through a Fourier transform lens. The angular multiplexing method for multiple holograms of database is achieved by controlling the reference directions with a step motor. Experimental results show that the proposed system can be used for fingerprint recognition.

The new image factorization algorithm, the k-factor method, is applied to the problem of shadow removal. Unlike the other technique, PCNN factorization, the k-factor method will completely remove shadows. It has several features that allow automatic or manual control that tailor it to specific applications. Unlike histogram equalization which globally remaps rather than remove shadows, the k-factor method eliminates regional shadows without remapping other regions. The algorithm is presented, numerous examples of its performance shown, and future plans discussed.

The resolution of many viewing and staring systems is often restricted by the spatial limited resolution of its sensing device rather than by diffraction limits related to the optical system. This spatial resolution of the sensor when limited by the pixels' dimensions is coined hereby 'geometrical resolution.' In this paper, we suggest a technique for overcoming this limit, thus obtaining 'geometrical super resolution.' The proposed approach is based on time sub pixel scanning procedure followed by a Gabor transform. In order to improve the obtained results a special pre-designed mask is attached to the sensing plane. When the sensor is placed on a fixed platform, the set of sub-pixel shifted images must be obtained by deliberate sub pixel scanning procedures. However, in cases when the sensor is placed on a vibrating platform, the vibration can be used to obtain the desired images set by capturing several images (while vibrating) and applying a registration algorithm.

With the emergence of a global economy, companies are more than ever pressured for improved efficiency. Int he transportation industry there is a growing need for better tracking of the status of containers in transit. This would lead to improved handling operation, reduce the number of errors, increase the throughput and enable the use of electronic data interchange (EDI). As electronic tags are not generalized in this industry, containers identification must rely on optical character recognition of the codes printed on the containers. OCR has been one of the first applications envisaged for optical correlation technologies as a result of their high-speed direct detection and identification capabilities. Until now though, most of the work in this area had been performed on computer-generated symbols. Field applications however, must cope with varying symbol fonts and sizes, colors and backgrounds, illumination levels, etc. Environmental variables such as dust, dirt and rust must also be accounted for. Together, these variables lead to a hard-to- solve problem. This paper presents INO's optical correlator and discusses the methods used to generate the identification vectors from which the OCR classification is achieved. It is shown that good results can be obtained on gray-scale real- life images when a multiple composite-filters strategy combined to an innovative classification method.

Using high speed imaging techniques, the gall bladder stone immersed in liquid is detected and identified. The detection of the shock waves induced by laser power is reached by using interferometry technique. Using gall bladder and tissue images obtained by ultra-fast photography and time resolved laser fluorescence the correlation of correlation is performed. The tissue image is used to perform the correlation filter. Hence lower correlation output is used for firing of the laser power.

This paper presents a methodology to evaluate and compare different algorithms for general pattern recognition problems. An abstract framework for considering the selection of algorithms is presented and used for the character recognition problem. It includes a problem space, an algorithm space, a subset of algorithms from which selection is to be made, and measures of the performance of algorithms. Numerical experiments are run to evaluate some statistical algorithms for the character recognition problem.

We use the cubic spline interpolation instead of the linear interpolation of the feature space trajectories for 3-D object recognition from its 2-D perspective views. The accuracy of the interpolated trajectories to the testing views is greatly enhanced by this change. The experiments are on the real-world IR images of military vehicles. Edge-based approach is used to reduce the effect of the changes in infrared image intensity distribution due to temperature changes in some running parts of vehicles.

To continue the study we reported last year in this conference, we would like to present here the theoretical origin of our design of this super-fast learning, neural network pattern recognition system. As we published in the last few years, a one-layered, hard-limited perceptron can be used to classify analog pattern vectors if the latter satisfy the PLI condition. For most pattern recognition applications, this condition should be satisfied. When this condition is satisfied, then an automatic feature extraction scheme can be derived using some N-dimension Euclidean geometry theories. This automatic scheme will automatically extract the most distinguished parts of the N-vectors used in the training. It selects the feature vectors automatically according to the descending order of the volumes of the parallelepiped spanned by these sub-vectors. Theoretical derivation and numerical examples revealing the physical nature of this process and its effect in optimizing the robustness of this novel pattern recognition system will be reported in detail.

We have recently demonstrated a compact, high speed, gray- scale optical correlator for target detection. The capability of the direct gray-scale scene input and the gray-scale (real- valued) filter modulation enables us to implement a near- theoretical optimal filter on the optical correlator. This paper describes filter synthesizing algorithm for detecting targets in cluttered background input scene and the projection from the complex filter version to the real version for implementation on the gray-scale optical correlator. It is based on optimal-tradeoff MACH filter. It was found that using an appropriate simulated noise image to substitute the commonly used white noise in the filter design procedure is a very effective way to suppress clutter noise while maintain high tolerance for distortion. Both simulation and experimental results are provided.

Jet Propulsion Laboratory (JPL) has recently developed a camcorder-sized grayscale optical correlator (GOC) for real- time automatic target recognition applications. Key features of this GOC include: a grayscale input SLM to accommodate direct interface with the input imaging sensor, a real-valued (bipolar-amplitude) filter spatial light modulator to enable use of a MACH (Maximum Average Correlation Height) composite correlation filter algorithm, compact and portable. This GOC architecture has greatly improved the system complexity by removing the need of preprocess (binarization) the input, and the powerful MACH composite filter has greatly reduced the number of filter templates. Updating of the GOC system will be described in this paper, a recent real-time field demonstration for target recognition and tracking at Mojave Desert, CA will also be reported.

In this paper we present a variation of the polynomial correlation filter (PCF) called constrained correlation polynomial filter (CPCF). We investigate the performance of this filter in the presence of noise. The peak-to-sidelobe ratio measure and the public MSTAR images database are used for evaluation. The effect of different terms in the polynomial filter is examined by simulation. Then, we introduce a theoretical framework called energy projection to predict the effectiveness of different terms in the CPCF.

Poor performance in key components has limited the development of optical target recognition systems. New components are now available, however, that exhibit size and speed characteristics compatible with image processing applications. The U.S. Army Aviation & Missile Command has recently begun a program to exploit the new devices for both homing and imagery analysis applications. New architectures and algorithms which exploit the modulation characteristics of the new devices are being developed and tested. Part of the program is also aimed at funding improvements in the devices to better meet the requirements for optical target recognition. This paper will discuss the overall program, specific devices under consideration, and present the architectures and algorithms under development. Any experimental results available at the time will also be presented, with more detailed results to be presented at a later conference.