Two techniques for recognizing three-dimensional(3D) objects based on passive and active optical sensing followed by numerical correlation are presented. One technique uses passive sensing of 3D object based on integral imaging and the other technique uses active sensing based on digital holography. In both techniques, the 3D data is stored in two-dimensional(2D) form as digital format and then the detected 3D information is used for recognizing 3D object based on 2D correlation techniques or neural network. Experimental results in both techniques are presented.

We consider rejection and classification tests on the MSTAR (moving and stationary target acquisition and recognition) public database. We follow a benchmark procedure, which involves classification of three object classes and rejection of two confusers. This problem is difficult, since MSTAR images are specular and each target has a full 360° aspect angle range. In addition, a classifier should be able to handle object variants and depression angle differences between the training and test sets. We employ a new support vector representation and discrimination machine (SVRDM) for its excellent rejection-classification capability. A new simple registration method is used. Test results are presented and compared with those of other algorithms. The proposed method was also applied to clutter rejection and produced perfect rejection scores.

JPL has developed a compact portable 512 x 512 Grayscale Optical Correlator [1-4] by integrating a pair of 512 x 512 Ferroelectric Spatial Light Modulator (FLCSLM), a red diode laser, Fourier optics, a CMOS photodetector array. The system is designed to operate at the maximum speed of 1000 frames per second. A FPGA card was custom programmed to perform peak-detection post-processing to accommodate the system throughput rate. Custom mechanical mounting brackets were designed miniaturized the optics head of the GOC into a 6” x 3.5” x 2” volume. The device driver HW/SW is installed in a customized PC. The GOC system’s portability has been demonstrated by shipping it to various locations for target recognition testing.

The Generalized Phase Contrast (GPC) method enables virtually loss free conversion of spatial phase patterns to highly efficient light intensity distributions. The GPC-method has been used in a number of applications requiring parallel light-beam encoding such as in advanced user-controlled optical micro-manipulation, wavefront sensing and generation for common-path interferometry and adaptive optics, optical phase-only encryption and integrated micro-optical implementations. In this work, we will outline the concept for a GPC-platform for advanced and user-interactive manipulation of fluid-borne colloidal structures with state-of-the-art controllability and versatility. Real-time reconfigurable light patterns with sub-micron accuracy are obtained from a direct map of phase patterns addressed on a programmable phase-only spatial light modulator device. A graphical user interface enables real-time, interactive and arbitrary control over the dynamics and geometry of synthesized light patterns. Arrays of GPC-generated counterpropagating light fields provides for multi-particle trapping and manipulation in three dimensions by incorporating a spatially-addressable polarization modulator where individual relative strengths of orthogonally-polarized beam-pairs are independently adjustable. The result is a system with fully independent control of a plurality of particles throughout all spatial dimensions and in real time, hence 4D optical multi-beam manipulation.

In this paper, we investigated automatic target detection and classification of low and high contrast targets present in unknown forward looking infrared (FLIR) image sequence. The detection algorithm, based on morphology based preprocessing, acts as a prescreener that selects possible candidate target regions, comprising both true targets and false alarms and places expected target-sized marker to those preselected regions. The application of simple non-linear grayscale operations in the proposed detection algorithm leads to real-time implementations. By considering the known target and background specific attributes, extracted from the training samples, the clutter rejection module discriminates between true target and false alarms previously identified by the detection algorithm. Two approaches are employed for object classification where one uses local features of the image and the other uses template matching technique such as image correlation. For the first approach, to extract features, we employed two methods - nonlinear filtering for texture energy measurement and wavelet decomposition by expending Daubechies high and low pass filter coefficients. Then for classification, a neural network based classifier is used. In the second approach minimax distance transform correlation filter (MDTCF) is applied that minimizes the average squared distance from the filtered true-class training images to a filtered reference image while maximizing the mean squared distance (MSD) of the filtered false-class training images to this filtered reference image. Then classification is performed using the squared distance of a filtered test image to the chosen filtered reference image. The performance of the proposed technique is analyzed for i) neural network with nonlinear texture filtering, ii) neural network with wavelet decomposition and iii) correlation filtering. Preliminary results indicate that the proposed detection algorithms can locate both hot and cold targets from cluttered background. In addition, the clutter rejecters are capable of maintaining a low false alarm rate and excellent discrimination competence. The performance of the proposed techniques has been tested with real life FLIR imagery supplied by the Army Missile Command (AMCOM).

We present new imaging techniques based on computed restoration from holographic wavefront detection. This technique allows for unlimited resolution, optical and turbulent aberrations can be corrected. The relative movement of target and detector is computed and input to the image restoration algorithm. We discuss how this imagery can be extended to active referenceless holography or even to passive incoherent image.

A method of detecting target objects in still images despite any kind of geometrical distortion is demonstrated. Two existing techniques are combined, each one capable of creating invariance to various types of distortion of the target object. A Maximum Average Correlation Height (MACH) filter is used to create invariance to orientation and gives good tolerance to background clutter and noise. A log r-θ mapping is employed to give invariance to in-plane rotation and scale by transforming rotation and scale variations of the target object into vertical and horizontal shifts. The MACH filter is trained on the log r-θ map of the target for a range of orientations and applied sequentially over regions of interest in the input image. Areas producing a strong correlation response can then be used to determine the position, in-plane rotation and scale of the target objects in the scene.

JPL is developing an Advanced Autonomous Target Recognition (AATR) technology to significantly reduce broad area search workload for imagery analysts. One of the algorithms to be delivered, as part of JPL ATR Development and Evaluation (JADE) project, is the OT-MACH based ATR algorithm software package for grayscale optical correlator. In this paper we describe the basic features and functions of the software package as currently implemented. Automation of filter synthesis and test for GOC, particularly the automation of OT-MACH parameter optimization, is discussed.

Composite correlation filters have been demonstrated in many automatic target recognition (ATR) applications because of their ability for class recognition and distortion-tolerance with shift invariance. Both the optimal tradeoff synthetic discriminant function (OTSDF) filters and optimal tradeoff distance classifier correlation filter (OTDCCF) approaches use parameters to combine multiple characteristics. Usually a set of filters is grouped into a bank for recognizing multiple targets across multiple geometric distortions. We extend these approaches to use independent tradeoff parameters in the filter synthesis for each class and grouping bin to improve classification. A method for determining the extended parameters is presented. Test results using the public SAR imagery MSTAR database are shown.

A face recognition system that functions in the presence of illumination variations is presented. It is based on the minimum noise and correlation energy (MINACE) filter. A separate MINACE filter is synthesized for each person using an automated filter-synthesis algorithm that uses a training set of illumination differences of that person and a validation set of a few faces of other persons to select the MINACE filter parameter c. The MINACE filter for each person is a combination of training images of only that person; no false-class training is done. Different formulations of the MINACE filter and the use of two different correlation plane metrics: correlation peak value and peak-to-correlation plane energy ratio (PCER), are examined. Performance results for face verification and identification are presented using images from the CMU Pose, Illumination, and Expression (PIE) database. All training and test set images are registered to remove tilt bias and scale variations. To evaluate the face verification and identification systems, a set of impostor images (non-database faces) is used to obtain false alarm scores (P_FA).

A new technique using a combination of maximum average correlation height (MACH) filter and polynomial distance classifier correlation filter (PDCCF) for simultaneous detection and classification of single/multiple identical and dissimilar targets is proposed in this paper. In this technique, a MACH filter is formulated for each desired target class from the training images of the corresponding target with expected size and orientation variations such that the size of the filter is the same as the input scene. Then a multi-class PDCCF is formulated from the training images of all target classes such that the size of the filter is the same as the expected targets. For real time applications, the input scene is first correlated with all MACH filters and the correlation outputs are combined. The regions of interest (ROI) containing the probable targets are selected from the input scene based on the ROIs with higher correlation peak values in the combined correlation output. The PDCCF filter is then applied to these ROIs to identify target types and reject clutters and/or backgrounds. To increase the robustness of the proposed technique, multiple filters are formulated for multiple ranges of target size and/or orientation variations. This two-stage system is faster and yields more accurate results compared to the existing three-stage system, which involves wide area prescreening, detection using MACH filters, and classification using distance classifier correlation filter. The simulation results using real life imagery show that the proposed technique can detect and classify the desired targets with higher efficiency irrespective of their distortion or the number of targets present in the input scene, when compared to the alternate techniques.

We describe a pose-invariant facial verification system for cooperative subjects for use in a controlled entry system. Facial images in several poses are acquired for template correlation to construct synthetic discriminate binary phase only filters (SDBPOF). SDBPOF is used to compensate for variances such as head rotation and tilt. Correlation accuracy is determined by SDBPOF construction. Subject identification based on facial appearance was performed with a digitally simulated 4-f optoelectronic correlator comparing performance between binary phase only filters (BPOFs) and SDBPOFs. Recognition accuracy and false positive rates were determined for each mode of operation.

Detection is one of the most formidable problems in automatic target recognition, since it involves locating multiple classes of targets of interest with distortions present in cluttered scenes. Fast and efficient algorithms are needed for detection, since in detection we need to analyze every local region of large image scenes. Minimum noise and correlation energy (MINACE) filters are attractive distortion-invariant filters (DIFs); we consider MINACE filter use in detection, since they provide sharp correlation peak values for targets and overcome the effect of aspect view distortions in the input data. Most prior work on DIFs considered classification, not detection. MINACE filters seem to require fewer filters than do other DIFs, and they recognize objects with aspect views different by 15° from those present in the training set. They are also shift-invariant and require only a few filters to handle detection of multiple target classes. We test our improved MINACE filters to detect 8 classes of objects in an infrared (IR) database with a ±90° range of aspect views. Initial test results are excellent with only 3 filters needed and very low false alarm rates.

JPL has recently developed an all-optical Joint Transform Optical Correlator (OJTC) utilizing a high-speed rewritable holographic photopolymer film (HPF) as the key real-time holographic read/write medium. The high resolution HPF has enabled the use of an off-axis holographic recording scheme that has completely eliminated the zero-order crosstalk plaguing most of the state-of-the-art OJTC systems with an on-axis recording scheme. The high sensitivity and fast erasure capability of the HPF film has made possible the real-time updatable target recognition performance of the OJTC. The OJTC system architecture, optical implementation and experimental results demonstrating fingerprint recognition are described in this paper.

In this paper we propose a new operational mechanism for an optically addressed deformable mirror device. The device consists of a pixilated metallized membrane mirror supported above an optically addressed photoconductive substrate. A conductive transparent conductive electrode is deposited on the backside of the substrate. A DC bias is applied between the membrane and the back electrode of the device accompanied with very high frequency modulated light. The membrane is deformed when light is shone from the backside of the device. This occurs due to impedance and bias redistribution between the two cascaded impedances.

The fringe-adjusted joint transform correlator (JTC) technique has been found to yield substantially better correlation performance than alternate JTC techniques. Since the fringe-adjusted JTC (FJTC) is sensitive to scale and rotation variations, a synthetic discriminant function (SDF) based FJTC was proposed to realize scale and rotation invariant pattern recognition system via computer simulation. In this paper, optoelectronic implementation of the scale and rotation invariant pattern recognition using SDF based FJTC has been tested for both binary and gray level images. The experimental results obtained are in close agreement with the simulation results obtained earlier.

JPL and BNS Inc. are jointly developing a compact, low mass, Electro-Optic Imaging Fourier Transform Spectrometer (E-O IFTS) for hyperspectral imaging applications [6]. The spectral region of this spectrometer is in the near IR spectral band of 1 - 2.5 μm (1000 - 4000 cm^-1) to allow high-resolution, high-speed hyperspectral imaging applications. The specific applications for NASA’s missions will focus on the measurement of a large number of different atmospheric gases simultaneously in the same airmass. Due to the use of a combination of birefringent phase retarders (YVO₄) and multiple achromatic phase switches to achieve phase delay, this spectrometer is capable of hyperspectral measurements similar to that of the conventional Fourier transform spectrometer but without any moving parts. In this paper, the principle of operations, system architecture and recent technical progress will be presented.

An innovative multispectral pattern recognition system, consisting of a surface plasmon tunable filter (SPTF) and a compact Grayscale Optical Correlator (GOC) is under development at JPL. The surface plasmon tunable filter is capable of rapidly acquiring narrowband input imagery sequence through the visible spectral band (400nm-1000nm) and feed into the cascading GOC for parallel target recognition. The combined system will be useful for one-pass hyperspectral imaging and pattern recognition.

A new processing of spectra for pattern recognition was created in order to detect explosives. Partial Least Squares (PLS) was used to create vector for recognition and those were using for discriminant analysis. PLS was adjusted with a discriminant function. IR spectra of TATP, DNT and TNT traces in air were recorded. Spectra of free air of those explosives were measured. NIR and MIR regions were studied and were used for PLS vector. NIR region is statistically significant. Two PLS were necessary for good discrimination for those explosives.

There have been renewed interests in utilizing imaging polarimetry in target detection and discrimination. A greater demand is the need for a sensor capable of discriminating between real military targets and decoys in the battlefield deployments. This paper demonstrates the potential application of passive visible imaging polarimetry in discriminating real targets from identical (same paint type, surface structure, and color) decoys, based on their composite materials. Target material made of steel is compared to three different decoy materials (wood, ceramic, and cardboard) are considered in this study.

The parallel processing capability and adaptive filtering features of dynamic neural networks offer highly efficient feature extraction and enhancement capability for fingerprint images. The most important aspect of the fingerprint enhancement is the extraction of relevant details with respect to distributed complex features. For this purpose, an efficient dynamic neural filtering technique has been proposed in this paper. After the enhancement process, fingerprint identification is/has been achieved using joint transform correlation (JTC) algorithm. Since the fringe-adjusted JTC algorithm has been found to yield significantly better correlation output compared to alternate JTCs, we used it in this study. The identification test results are presented to verify the effectiveness of the proposed enhancement and identification algorithms.

In a set of preprocessed N-dimension, analog pattern vectors {U_m, m=1 to M, each U_m represents a distinct pattern}, if N>>M, then a one-layered sign-function neural network (OLNN) is sufficient to do a very robust, yet very accurate, noniterative-learning of all patterns. After the learning is done, the OLNN will make an accurate identification on an untrained test pattern even when the test pattern is varying within a certain dynamic range of a particular standard pattern learned during the noniterative learning process. The analytical foundation for making this dynamic neural network pattern recognition possible is the following. If we know that a standard pattern U_m will vary gradually among K boundary patterns U_m1 to U_mk, then we can train the neural network noniteratively to learn JUST THE BOUNDARY vectors {U_mi, i=1 to k} for each pattern U_m. Then, due to a distinctive property of noniterative learning, for a test input pattern U_t equal to any graduate changes within the boundaries (i.e., U_t = any CONVEX combination of the boundary set {U_mi, i=1 to k, m fixed.}), the OLNN can still automatically recognize this changed pattern even though all these gradually changed pattern are NOT learned step by step in the noniterative learning.

Zernike moments are one of the most effective orthogonal, rotation-invariant moments in continuous space. Unfortunately, the digitization process necessary for use with digital imagery results in compromised orthogonality. In this work, we introduce improved digital Zernike moments that exhibit much better orthogonality, while preserving their inherent invariance to rotation. We then propose a novel pattern recognition algorithm that is based on the improved digital Zernike moments. With the improved orthogonality, targets can be represented by fewer moments, thus minimizing computational complexity. Additionally, the rotation invariance enables our algorithm to recognize targets with arbitrary orientation. Because our algorithm eliminates the segmentation step that is typically applied in other techniques, it is better suited to low-quality imagery. Simulations on real images demonstrate these aspects of the proposed algorithm.

We discuss a novel control methodology for power management in heterogeneous distributed sensor networks. Many algorithms for resource management in sensor networks require a comprehensive model of the external environment and the sensor network system, and are rule-based; this restricts their use in dynamic environments. We present an event-based control optimization formulation of the resource management problem and discuss a method to adaptively change desired system performance of the sensor network in response to events. This functionality is critical in field-deployable sensor networks where the available energy is extremely limited. This limitation disallows continuous operation as a very expensive option and necessitates system adaptation as a means to extend operational lifetime in the face of dynamic external events. We show results on synthetic sensor networks where only partially accurate information about the external world and the sensing system is available and illustrate the efficacy of the control algorithm in handling dynamic events with guaranteed minimum system lifespan via efficient usage of energy resources. We show that the control algorithm makes effective control decisions about the use of energy and storage resources with varying sensor reliabilities.

An incoherent correlator system for star pattern recognition and tracking was described by Kouris et al. in SPIE Vol. 4734, as part of the 2002 OPR XIII Conference Session. The experimental demonstration described utilized small but discrete bulk mirrors as the specialised matched filtering elements. The replacement of the discrete bulk mirrors by a matrix of MEMS mirrors is desirable, especially in space applications, due to the light weight, compact size, low power consumption and precise patterning of MEMS structures. We describe here the relation of MEMS mirror parameters to star tracker performance. Two approaches to fabricate a two dimensional array of large area, high tilt angle and fast switching MEMS mirrors with orthogonal tilt axes are discussed, as required in the star tracker application. A preliminary design and test results of the MEMS array are presented.

JPL is developing a high resolution (512 pixel x 512 pixel), high-speed (1000 frames/sec), compact Automatic Target Recognition (ATR) processor for onboard target detection, identification and tracking. This ATR processor consists of a compact Grayscale Optical Correlator (GOC) for parallel wide area target-of-interest (TOI) detection and a hardware based self-learning neural network (NN) for target identification and adaptive monitoring. This processor can be tailored to meet specific system requirements for many ATR applications. Development includes simulation of key components in software including GOC simulation and NN simulation. Both simulation tools are discussed and demonstrated.

In this paper we present a multiple reference optoelectronic joint transfer correlator (JTC) in which the reference images are phase encoded using a pseudorandom phase sequence. The phase encoding is used to remove extraneous peaks from the correlation plane as well as improve the spatially efficiency of the JTC system. The reference images are phase encoded apriori making the system fast for real-time feedback. The optoelectronic joint transform correlator (JTC) is presented and analyzed as an effective optoelectronic correlator. The architecture of the proposed system is presented.

A new three-dimensional (3D) color pattern recognition technique, utilizing the concept of fringe-adjusted joint transform correlator (JTC) and CIELAB color space, is proposed in this paper. The proposed technique yields better discrimination capability, sharper and stronger correlation peak intensity, compared to classical joint transform correlator with conventional red-green-blue (RGB) components. Simulation results verify the robustness of the proposed technique.

In previous work we introduced a new metric for the image resolution volume that couples the spatial resolution with the range resolution. We showed from a quantum noise consideration that there is a constant volume resolution and that one can trade-off spatial resolution at the expense of range resolution, and vise-a-versa. This theory was developed for a heterodyne LADAR system. In this paper we extend our previous heterodyne LADAR system theory to develop a image resolution volume metric for time-of-flight LADAR system, where device parameters such as optical amplifier noise are included.

During the last decade we have extended the implementation of companding techniques in communication theory to apply to improve image processing in several optical systems by using implementations using nonlinear optical media. In this paper we introduce a photorefractive two-beam-coupling deconvolution using spatially-variable dynamic spectral compression. Resolution recovery of blurred noisy images is demonstrated for several different type of image blur.

We introduce for the first time a novel and unique algorithm for a generalized form for a minimum mean-square-error image processing filter. This algorithm can be used to recognize or retrieve an image that is not only partially obscured by a constant disjoint background, but is also simultaneously blurred and overlaid with additive gaussian noise. Although this algorithm can be applied to many general filter forms that have never been considered before, we test the performance of this filter in four novel obscured-version operating modes: three recognition modes and one retrieval mode. These tests included varying the levels of the background illumination and of the additive white noise, as well as varying the amount of obscuration on the image. Our simulation results show that it is possible to recognize or retrieve images that are as much as 90% obscured, as well as blurred and noisy with a signal-to-noise ratio of 0.1. We also show that the background illumination of the obscuring object improves the performance of the filter in both its recognition or retrieval modes. This work should be a significant advance in the pattern recognition area for both automatic target recognition and machine vision.

Face recognition based on independent component analysis (ICA) has emerged as a popular approach for face recognition application. In this paper we present a comparison between various optoelectronic face recognition techniques and ICA based face recognition. Computer simulations are used to study the effectiveness of the fastICA algorithm in recognizing facial images with a high level of three-dimensional (3-D) distortion. Results are then compared to various distortion-invariant optoelectronic face recognition algorithms such as synthetic discriminant functions (SDF), projection-slice SDF, optical correlator based neural networks, and pose estimation based correlation.

This paper describes the basic N-dimension geometrical meaning of the noniterative neural network and the geometrical derivation of one of the most important properties of this neural network: The optimum robustness in the recognition of the untrained patterns. Based on this concept of optimum robustness, a novel automatic feature extraction system is derived. The predicted optimum robustness and the ultra-fast learning speed of this novel system are then verified experimentally. This paper concentrates at the geometrical derivations of this novel neural system design.

We present a new class of optical-electronic reconfigurable image processing systems, controlled by the parameters of the input image. The systems use multiprocessors and are distributed. We describe two different optical processors, each performs the Fourier transform and correlation with an adaptive filter or with a fixed set of filters.

It is proposed a new optical electronic approach for effective, simple and non expensive testing of the materials. An optical correlator is used for high speed features extraction, which characterize the distribution of the informational important elements in the crystallographic image. The digital “portrait” of the analyzed material is constructed which is compared with the set of the standard “portraits” on the base of which the level of the quality of the material is determined. The method permits to automate the process of the crystallographic images analyses and to increase the reliability of the results.

We have demonstrated an optical novelty filter based on the two-beam coupling effect in photorefractive polymers. The photorefractive polymer composition was optimized for response time and two-beam coupling gain by changing the ionization potential and polarizability of various components. In this study, a photorefractive polymer composition was simultaneously optimized for response time and gain, and employed as a key element in a two-beam coupling novelty filter with a high contrast ratio and a limiting frequency of 14Hz, considerably higher than any previously reported in a two-beam coupling photorefractive novelty filter.

Prior studies of multimodal biometric fusion have shown that it can improve performance over use of a single unimodal biometric. The well-known multimodal methods do not consider the quality information of the data used when combining the results from different matchers. In addition to examining these well-known fusion methods, we introduce novel methods of fusion which combine face and fingerprint biometric results using fingerprint data quality information. We show that multimodal biometric fusion using data quality information outperforms standard multimodal results and unimodal systems.