Face recognition with both pose and illumination variations is considered. In addition, we consider the ability of the classifier to reject non-member or imposter face inputs; most prior work has not addressed this. A new SVRDM support vector representation and discrimination machine classifier is proposed and initial face recognition-rejection results are presented.

JPL is developing a portable 512 x 512 Grayscale Optical Correlator (GOC) system for target data mining and identification applications. This GOC system will utilized a pair of 512 x 512 Ferroelectric Liquid Crystal Spatial Light Modulator (FLCSLM) to achieve 1000 frames/sec data throughput. Primary system design issues including: optics design to achieve compact system volume with fine tuning capability, photodetector array with onboard post-processing for peak detection and target identification. These issues and corresponding solutions will be discussed.

We present an optical method for information watermarking of 3D objects using digital holography and authenticating the hidden image. A hidden image is embedded using double phase encoding in a phase shift digital hologram of the 3D object. The watermarked hologram is decoded to reconstruct the hidden image and the 3D object. We use either the entire hologram or a part of it to decode the hidden image. The recovered hidden image is authenticated using non-linear correlation. Experiments are presented to illustrate the ability to recover both the 3D object and the decoded hidden image.

Correlation filters can be very effective for object recognition. However, these filters may become too computationally expensive when applied to large images, or large numbers of images, because they require Fourier transforms between the spatial and frequency domains. This paper makes the simplifying assumption of single object recognition and presents an algorithm designed to reduce complexity of computation and/or storage. The algorithm derives a frequency domain match metric as opposed to the standard approach of using the spatial correlation plane. The performance of the efficient algorithm is compared to that of the standard correlation filter algorithm, for both accuracy and computational requirements.

Presented in this paper is a fringe-adjusted joint transform correlator (FJTC) based invariant target tracking of forwar looking infra-red (FLIR) image sequences. The proposed FJTC based tracking approach employed a modified synthetic discriminant function (SDF) concept together with an efficient camera motion compensation technique to accommodate the problem of target signature variation due to in-plane/out-of-plane rotations, scale variations, noise, and bad frames. The proposed technique can track small objects comprising of only a few pixels and is capable of compensating the high ego-motion of the sensor. The robustness of the proposed technique is demonstrated with computer simulation performed on sequences of eal life FLIR imagery taken from an airborne moving platform.

Previously we have described a hybrid optical neural network (HONN) filter. The filter is synthesised employing an artificial neural network technique that generates a non-linear interpolation of the intermediate train set poses of the training-set objects but maintains linear shift-invariance which allows potential implementation within a linear optical correlator type architecture. In this paper, we remove the constraints imposed on the filter’s output correlation peak height from the constraint matrix of the synthetic discriminant function used to create the composite filter. We examine the U-HONN filter’s detectability, peak sharpness, within-class distortion range, discrimination ability between an in-class and out-of-class object and the filter’s tolerance to clutter. We assess the behaviour of the U-HONN filter in an open area surveillance application. The filter demonstrates good object detection abilities within cluttered scenes, keeping good quality correlation peak sharpness and detectability throughout all the sets of tests. Thus the U-HONN filter is able to detect and accurately classify the in-class object within different background scenes at intermediate angles to the train-set poses.

Several metrics for quantifying the performance of fringe-adjusted joint transform correlator (JTC) technique are investigated in this paper. The criteria used for measuring the performance of fringe-adjusted JTC include peak sharpness, signal-to-clutter measure, distortion invariance, signal-to-noise ratio and a new metric called peak-to-background correlation energy is proposed in this paper. These metrics are used to estimate the reliability of signal detection in the input scene with respect to clutter, noise and other associated distortions. Detailed analysis and simulation results for quantifying the performance of fringe-adjusted JTC are presented.

A modified phase-encoded fringe-adjusted joint transform correlation technique is proposed for multiple target detection, where two joint power spectrums (JPS) are formulated utilizing a random phase mask and phase shifted random phase mask to the reference image separately. The final JPS is the difference between the phase encoded and shifted phase encoded JPS which is multiplied by the phase mask before applying the inverse Fourier transform to yield the correlation output. This technique ensures better utilization of the input/output plane space bandwidth product by yielding one delta function like correlation peak for each desired target object and no peak for non-target objects. The proposed technique can effectively detect any number of targets from noise free or noisy input scenes without changing the system parameters and without any degradation of performance. Computer simulation results verify the performance of the proposed technique.

An Optical Processing for the Mining and Identification of Targets (OPMIT) system is being proposed to significantly reduce broad area search workload for NIMA imagery analysts. Central to the system is a Grayscale Optical Correlator (GOC), developed by JPL in recent years. In this paper we discuss some preliminary development of an important system component - the filter management module - that is critical for the success of GOC operation. The emphasis is on the streamlining the OT-MACH filter synthesis/testing procedure for effective and efficient filter design while maintaining filter performance.

The detection and discrimination of targets in infrared imagery has been a challenging problem due to the variability of the target and clutter (background) signatures. In this paper we discuss the application of a novel quadratic filtering method using missile seeker infrared closing sequences. Image filtering techniques are well suited for target detection applications since they avoid the disadvantages of typical pixel-based detection schemes (such as segmentation and edge extraction). Another advantage is that the throughput complexity of the filtering approach, in the detection process, also does not vary with scene content. The performance of the proposed approach is assessed on several data sets, and the results are compared with that of previous linear filtering techniques. Since we can obtain the signature of some of the clutter “in-the-field” or during operation, we examine the impact of updating the filters to adapt to the clutter.

At the Optical Computing Lab in the Jet Propulsion Laboratory (JPL) a binary holographic data storage system was designed and tested with methods of recording and retrieving the binary information. Levels of error correction were introduced to the system including pixel averaging, thresholding, and parity checks. Errors were artificially introduced into the binary holographic data storage system and were monitored as a function of the defect area fraction, which showed a strong influence on data integrity. Average area fractions exceeding one quarter of the bit area caused unrecoverable errors. Efficient use of the available data density was discussed.

The fabrication and characterization of an optically addressable deformable mirror for spatial light modulator is described. Device operation utilizes an electrostatically driven pixellated aluminized polymeric membrane mirror supported above an optically controlled photoconductive GaAs substrate. A 5-μm thick grid of patterned photoresist supports the 2-μm thick aluminized Mylar membrane. A conductive ZnO layer is placed on the backside of the GaAs wafer. Similar devices were also fabricated with InP. A standard Michelson interferometer is used to measure mirror deformation data as a function of illumination, applied voltage and frequency. A simplified analysis of device operation is also presented.

For an optical correlator system, information is encoded in four dimensions (spatially, bits per pixel and time). This paper describes an optical correlator system that was built around 8-bit, 256 x 256 spatial light modulators capable of operating at four kilohertz frame rates. The full data throughput of the optical processor is not realized due to limitations with the correlation plane detector and post processor. However, continuous operation at nearly a kilohertz was demonstrated.

Previous field-deployable distributed sensing systems for health/biomedical applications and environmental sensing have been designed for data collection and data transmission at pre-set intervals, rather than for on-board processing These previous sensing systems lack autonomous capabilities, and have limited lifespans. We propose the use of an integrated machine learning architecture, with automated planning-scheduling and resource management capabilities that can be used for a variety of autonomous sensing applications with very limited computing, power, and bandwidth resources. We lay out general solutions for efficient processing in a multi-tiered (three-tier) machine learning framework that is suited for remote, mobile sensing systems. Novel dimensionality reduction techniques that are designed for classification are used to compress each individual sensor data and pass only relevant information to the mobile multisensor fusion module (second-tier). Statistical classifiers that are capable of handling missing/partial sensory data due to sensor failure or power loss are used to detect critical events and pass the information to the third tier (central server) for manual analysis and/or analysis by advanced pattern recognition techniques. Genetic optimisation algorithms are used to control the system in the presence of dynamic events, and also ensure that system requirements (i.e. minimum life of the system) are met. This tight integration of control optimisation and machine learning algorithms results in a highly efficient sensor network with intelligent decision making capabilities. The applicability of our technology in remote health monitoring and environmental monitoring is shown. Other uses of our solution are also discussed.

In a previous publication we introduced a new photorefractive four-wave mixing deconvolution, FWMD, image correction approach for achieving one-way image transmission through an aberrating medium. In this paper we extend our work to include additional image degradations and more test cases. We characterize the performance as a function of the input beam ratios for four metrics: signal-to-noise ratio (SNR), normalized mean-square-error (NMSE), edge restoration (ER), and the peak-to-total energy ratio (PTE). In our characterization we color-code the best beam-intensity ratio 2D region(s) for each of the above metrics. Test cases are simulated at the optimal values of the beam-intensity ratios.

In this paper, we introduce a new design concept of laser radar systems that combines both phase comparison and time-of-flight methods. We show from signal to noise ration considerations that there is a fundamental limit to the overall resolution in 3-D imaging range laser radar (LADAR). We introduce a new metric, volume of resolution (VOR), and we show from quantum noise considerations, that there is a maximum resolution volume, that can be achieved, for a given set of system parameters. Consequently, there is a direct tradeoff between range resolution and spatial resolution. Thus in a LADAR system, range resolution may be maximized at the expense of spatial image resolution and vice versa. We introduce resolution efficiency, η_r, as a new figure of merit for LADAR, that describes system resolution under the constraints of a specific design, compared to its optimal resolution performance derived from quantum noise considerations. We analyze how the resolution efficiency could be utilized to improve the resolution performance of a LADAR system. Our analysis could be extended to all LADAR systems, regardless of whether they are flash imaging or scanning laser systems.

JPL and BNS Inc. are jointly developing a compact, low mass, Electro-Optic Imaging Fourier Transform Spectrometer (E-O IFTS) for hyperspectral imaging applications. The spectral region of this spectrometer will be 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 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 experimental progress will be presented.

Hyperspectral and polarimetric data contain spectral response information that provides detailed descriptions of an object. These new sensor data are useful in automatic target recognition applications. However, such high-dimensional data introduce problems due to the curse of dimensionality, the need to reduce the number of features used to accommodate realistic small training set sizes, and the need to employ discriminatory features and still achieve good generalization (comparable training and test set performance). In this paper, we evaluate both hyperspectral and polarimetric feature sets and identify features useful for distinguishing targets from background. Various feature selection algorithms are assessed in terms of the goodness of the selected features and computation time. Our results show that (1) the integration of branch and bound algorithm and floating forward selection algorithm is promising for hyperspectral and polarimetric target detection applications; and (2) the combination of both hyperspectral and polarimetric features yields significantly better classification results than either hyperspectral or polarimetric features alone.

Eyetracking is one of the latest technologies that has shown potential in several areas including human-computer interaction for people with and without disabilities, and for noninvasive monitoring, detection, and even diagnosis of physiological and neurological problems in individuals. Current non-invasive eyetracking methods achieve a 30 Hz rate with possibly low accuracy in gaze estimation, that is insufficient for many applications. We propose a new non-invasive visual eyetracking system that is capable of operating at speeds as high as 6-12 KHz. A new CCD video camera and hardware architecture is used, and a novel fast image processing algorithm leverages specific features of the input CCD camera to yield a real-time eyetracking system. A field programmable gate array (FPGA) is used to control the CCD camera and execute the image processing operations. Initial results show the excellent performance of our system under severe head motion and low contrast conditions.

This paper presents digital image segmentation techniques and algorithms using variable threshold functions. The technique is successful to detect regions with different or poor light radiances and can be applied to images with occluded or noisy objects. The Variable threshold functions are derived from discrete time functions often used in digital control system design. The developed algorithms can be integrated on a single monolithic integrated circuit or implemented as an embedded system for real-time applications.

Stereo imagery has been a goal in optics research since the invention of the stereoscope in 1834. While the market has been inundated with displays of various types, sizes, and formats, no general purpose, easy to use, inexpensive method for the display of imagery in stereo has been developed. The benefits of stereo vision are numerous and quickly become apparent when attempting to perform simple tasks without the aid of stereo cues. The proliferation of remotely operated vehicles and indirect view applications has resulted in an increased need to see the operational environment in stereo. Numerous approaches to the display of stereo imagery have been demonstrated. Stereoscopic displays typically require the user to wear special headgear. Autostereoscopic displays, so named because they do not require the headgear, typically have tight limitations on the position of the viewer’s head. Previous papers have described the theoretical underpinnings for new type of stereoscopic displayed based on dual liquid crystal displays. The new display provides a stereo view without temporal or spatial multiplexing. This paper will present the results from experiments to characterize the display components and the resulting changes in the encoding algorithm.

We introduce novel optoelectronic target detection technique using polarization enhancement. Images correspond to elements of Stokes vector imagery are introduced as input scenes to the fringe-adjusted joint-transform correlator. Our results show excellent improvement in performance parameters. Both computer simulation and experimental results are presented in support of the proposed technique.

Any system that will aid the military pilot in directing fire or sensors or aid in the stabilization of imagery that is captured by the aircraft's sensors would substantially increase the pilot's effectiveness. Several optical techniques are described in this presentation for determining the orientation and position of a pilot's helmet, which is actually one of the first requirements in accomplishing the tasks just listed. These techniques are all passive in that they require no input from the pilot except perhaps for an initial calibration, which would likely be valid for a particular pilot for any subsequent flight. The problem of determining the position and orientation of a pilot's helmet can be thought of as determining the position and orientation of a vector that starts at the center of the pilot's head and points forward. If we take this vector to be of a fixed (arbitrary) length, the problem reduces to the determination of five independent variables. Several techniques are investigated that range from the very simple, direct view and map approach to complicated routines involving color mixing, optical correlation, time of flight measurements, intensity gradients, and fiber optic gyroscopes. All of these approaches work. The true goal of this investigation is to define the problem physically and mathematically, and to analyze all of the approaches and ultimately determine the advantages and disadvantages of each before much laboratory equipment has been dedicated and other expensive equipment purchased.

Feature detectors have been considered for the role of supplying additional information to a neural network tracker. The feature detector focuses on areas of the image with significant information. Basically, if a picture says a thousand words, the feature detectors are looking for the key phrases (keypoints). These keypoints are rotationally invariant and may be matched across frames. Application of these advanced feature detectors to the neural network tracking system at JPL has promising potential. As part of an ongoing program, an advanced feature detector was tested for augmentation of a neural network based tracker. The advance feature detector extended tracking periods in test sequences including aircraft tracking, rover tracking, and simulated Martian landing. Future directions of research are also discussed.

The problem of tracking the motion of a functional, electrically controlled hand is addressed in this paper. We show that by using skin color and hand segments, the motion of a human hand controlled by a functional electrical stimulation (FES) system can be recovered and predicted. The hand motion tracking and prediction information can then be used as feedback in a closed loop control system to improve the control of the hand motion. We show how skin color tracking can be used to segment the human hand regions and how the hand segments can be used to identify the hand moving positions and how the changes of hand segments between two frames can be used to track the hand orientations. We then show how the hand motion can be recovered based on the changes of the hand positions and orientations and how the motion can be simulated and predicted. The tracking and motion identification methods are verified by the presentation of experimental results obtained using the methods described in the paper.

In this paper we utilize the Karhunen-Love Transform to the Projection-Slice Synthetic Discriminant Function Filters, KLTPSDF to reduce the data set that represents each of the training images and to emphasize the subtle differences in each of the training images. These differences are encoded into the PSDF in order to improve the filter sensitivity to the recognition and identification of protein images formed from a cryo-electron microscopic imaging process. The PSDF has been shown to improve the performance of the SDF on specific images in previous papers. The protein structures found in cryo-electron microscopic imaging represent a class of objects that have low resolution and contrast and subtle variation. This poses a challenge in design of filters to recognize these structures due to false targets that often have the very similar characteristics as the protein structures. The incorporation of the KLT in forming the filter provides an optimal method of decorrelating images prior to their incorporation into the filter. We present our method of filter synthesis and the results of the application of this modified filter to a protein structure recognition problem.

An encryption/decryption scheme based on a new Phase-Contrast Technique, without the use of a phase-changing plate (phase dielectric dot) on the Fourier plane of a 4f optical correlator is proposed. The encryption of a gray level image is achieved by multiplying the phase distribution obtained directly from the gray level image by a random phase distribution. The encoding is obtained without any iterative calculation to generate the encrypted phase-only mask. The robustness of the encoding is assured by the non-linearities intrinsic to the phase-contrast method and the random phase distribution used in the encryption process. The advantage of this method is the easy scheme to recover the gray level information from the decrypted phase-only mask applying the proposed Zero-Order Phase-contrast Technique.

Classification and rejection tests were performed on the 10-class MSTAR database. To compare our performance, we first summarize relevant prior work. Shift-invariant magnitude Fourier transform (FT) features were used in the feature space trajectory (FST) classifier to classify the 10-class MSTAR data with variants and to reject confuser images and clutter chips. No prior work has addressed this. Implication of various SAR preprocessing on performance is addressed. In confuser rejection, 2 standard confusers (used in prior work) and 2 new confusers are addressed. We are the first to extend confuser rejection tests to 8 target classification with variants and 2 confuser rejection. Finally, clutter rejection while classifying the 10 targets is addressed.

A face verification system based on the use of a minimum noise and average correlation energy (MINACE) filter for each person is presented that functions with illumination variations present. A separate filter is used for each person; it is a combination of different training images of only that person. The system is tested using both unregistered and registered images from the CMU Pose, Illumination and Expression (PIE) database. The number of correct (P_C) and the number of false alarm (P_FA) scores are compared for the two cases. Rather than using the same parameters for the filter of each person, an automated iterative filter training and synthesis method is used. A validation set of several other faces is used to achieve parameter selection for good rejection performance. For filter-evaluation, all filters are tested against all images, but the same peak threshold is used for each filter to determine verification and rejection.

This paper analyzes the operation of a new optically addressed deformable mirror device for applications in adaptive optics and optical signal processing. Device operation utilizes a pixellated metallized polymeric membrane mirror supported above an optically addressed photoconductive substrate. A grid of patterned photoresist supports the metallized membrane. A conductive ZnO layer is placed on the backside of the substrate. The device operates as an impedance distribution between two cascaded impedances between the deformable membrane and substrate and the substrate and back electrode. We develop a theoretical model to analyze the deformation as a function of the light intensity and electrical drive

Automatic target recognition (ATR) technique has been applied in both civil and military. In this paper, we present a new optical pattern recognition system for target recognition. This system includes synthetic discriminate function (SDF) based practical optimized filters for the 3-D targets, the Reference Filter Libs for high correlation SNR, the mapping between the input (object regions) and the output (correlation peaks), and neural networks (ANN) for final decision making. The Real-time optical target recognition is realized by temporal multiplexing technique with electronic addressing spatial light modulator. The experiment results show that the proposed OPR system is efficient and reliable.

Fourier holograms are commonly used for reference images storing in diffraction correlators with spatially coherent or spatially incoherent illumination. Kinoforms can be a real alternative to Fourier holograms in the correlators. The kinoform represents a computer-synthesized optical element which performs only a phase modulation of a light wave. The kinoform restores true intensity of the recorded image and random distribution of phase. Therefore, it can be utilized for storing reference images, first of all, in correlators with spatially incoherent illumination. The absence of carrier frequency reduces demanded number of pixels of the spatial light modulator being used. Since the kinoform provides reconstruction of reference image in zero diffraction order, requirement on monochromaticity of illumination are decreased as well. The diffraction correlator with the kinoform used as spatial frequency filter is considered. The 2-D acoustooptic deflector was employed to form input images in real time by monochromatic spatially incoherent light. The reference images were recorded on the commercially available kinoforms. The input and reference images were of 256×256 pixels and 200×200 pixels respectively. Since input images were consisted of approximately 400 pixels with non-zero brightness, the image update frequency was gained at 200 Hz. The experimental setup and experimental results on images recognition are presented.

The theory of designing the optical-electronic image processing computer systems has been presented. A model of parallel image processing system has been considered, that is based on the principle of function decomposition. A structure model of computer system has been examined, that is a conveyor of parallel processors. The implementation possibilities of different image processing operations by optical and electronic computer means have been analyzed. Evaluation of time outlay in the system, while processing an image or a series of them has been made. There have been exposed the dependence of image processing time from conveyor length change and the correlation of optical and electronic devices. The designing method of image processing systems in static mode has been described. The results of investigations of the influence of the median square deviation, of the processing time in the modules on the throughput capacity of the system under the different electronic and optical modules quantity are presented. The recommendations of increasing the system’s throughput capacity are formulated. On the bases to these recommendations, the system design method of image processing in the dynamic mode is elaborated.