We compare the performance of our SVRDM (support vector representation and discrimination machine, a new SVM classifier) to that of other popular classifiers on the moving and stationary target acquisition and recognition (MSTAR) synthetic aperture radar (SAR) database. We present new results for the 10-class MSTAR problem with confuser and clutter rejection. Much prior work on the 10-class database did not address confuser rejection. In our prior work [1], we presented our results on a benchmark three-class experiment with confusers to be rejected. In this paper, we extend results to the ten-class classification case with confuser and clutter rejection. Our SVRDM achieved perfect clutter rejection scores, but the clutter was not demanding. Energy-normalization, which was used in many prior algorithms, makes clutter chips similar to target chips and thus produces worse results. We do not energy-normalize data.

In this paper we present an information sensing system which integrates sensing and processing resulting in the direct collection of data which is relevant to the application. Broadly, integrated sensing and processing (ISP) considers algorithms that are integrated with the collection of data. That is, traditional sensor development tries to come up with the "best" sensor in terms of SNR, resolution, data rates, integration time, etc. and traditional algorithm development tasks might wish to optimize probability of detection, false alarm rate, class separability, etc. For a typical Automatic Target Recognition (ATR) problem, the goal of ISP is to field algorithms which "tell" the sensor what kind of data to collect next and the sensor alters its parameters to collect the "best" information in order that the algorithm performs optimally. We demonstrate an ISP system which utilizes a near Infrared (NIR) Hadamard multiplexing imaging sensor. This prototype sensor incorporates a digital mirror array (DMA) device in order to realize a Hadamard multiplexed imaging system. Specific Hadamard codes can be sent to the sensor to realize inner products of the underlying scene rather than the scene itself. The developed ISP algorithm uses an ATR metric to send codes to the sensor in order to collect only the information relevant to the ATR problem. The result is a multiple resolution hyperspectral cube with full resolution where targets are present and less than full resolution where there are no targets. Essentially, this is compressed sensing.

Recent progress on developing a Grayscale Optical Correlator (GOC) is described. The development efforts have been simultaneously focused on the adoption of a composite correlation filter algorithm and the development and integration of high-resolution, high-speed, miniature GOC hardware system. We have selected the Optimum Trade-off Maximum Average Correlation Height (OT-MACH) filter synthesis algorithm due to its great target recognition performance and its suitability for optical implementation in a real-valued Spatial Light Modulator (SLM). We have, to date, participated in the development of the high speed (1000 frames/sec) and high resolution (1024 pixel x 1024 pixel), small form factor (5 micron pixel pitch) Ferroelectric Liquid Crystal SLM. We have also developed a portable 512 x 512 GOC system and used it to various ATR applications such as target detection from surveillance data, sonar mine detection, etc. The status of our GOC hardware system and correlation filter synthesis status will be reported. Potential applications and system issues will also be discussed.

This paper describes a system for the detection of people moving in a video scene. The system is ultimately intended to allow tracking of the major parts of the human body. Objects are detected utilising a background subtraction method which is based on the movements of objects within the scene. The background removal algorithm is designed in a way that allows the background to be constantly updated automatically, allowing it to be used over a long period of time. Several methods for improving the outcome from the background removal algorithm are used which include addressing problems caused by variable shading. Mathematical morphology techniques are subsequently employed in order to improve the segmentation achieved in each frame.

A novel one-dimensional (1D) fringe-adjusted joint transform (FJTC) correlation based technique is proposed for detecting very small targets involving only a few pixels in hyperspectral imagery. In this technique, spectral signatures from the unknown hyperspectral imagery are correlated with the reference signature using the 1D-FJTC technique. This technique can detect both single and/or multiple targets in constant time while accommodating the in-plane and out-of-plane distortion. Test results using real life hyperspectral image cube are presented to verify the effectiveness of the proposed technique.

A novel predictive controller framework is discussed for integrated smart sensing and resource management in a distributed intelligent sensor network with sensor nodes whose useful active lifetime is significantly shorter than the required lifetime of the sensor system. Past sensor network research has focused on security and communication, but largely ignored the overall dynamic resource management issue of such distributed systems. Our contribution is in integrated control optimization and resource management algorithms to ensure proper functioning of distributed sensors in extremely limited bandwidth, power, and storage. In this paper, we present a novel genetic algorithm (GA) for real-time system control and resource management that handles multiple conflicting objectives and constraints in the distributed system. We suggest using a GA-based predictive controller to give optimal actions that account for future states and future possible events while estimating the optimal control action policy. Our framework is suitable for dynamic environments where the desired system performance and resource usage changes dynamically while being constrained by limited amount of resources.

The minimum noise and correlation energy (MINACE) distortion-invariant filter (DIF) finds use in several applications such as face recognition, automatic target recognition (ATR), etc., in which one considers both true-class object classification and rejection of non-database objects (that are labeled as impostors in face recognition, and confusers in ATR). To solve the classification/rejection problem, we use at least one Minace filter per object class to be recognized. A separate Minace filter or a set of Minace filters is synthesized for each object class. The Minace parameter c trades-off distortion-tolerance (recognition) versus discrimination (impostor/confuser rejection) performance. We present our automated Minace filter-synthesis algorithm (auto-Minace) that selects the training set images to be included in the filter and selects the filter parameter c, so that the filter can achieve both good recognition and impostor/confuser and clutter rejection performance; this is achieved using a training and validation set. No impostor/confuser, clutter or test set data is present in the training or validation sets. The peak-to-correlation energy (PCE) ratio is used as the correlation plane metric in both filter synthesis and in tests, since it gives better results than use of the correlation peak value. We also address the use of the Minace filters in detection applications where the filter template is much smaller than the target scene. The use of circular versus linear correlations are addressed, circular correlations require less storage and fewer online computations.

A method of tracking objects in video sequences despite any kind of perspective distortion is demonstrated. Moving objects are initially segmented from the scene using a background subtraction method to minimize the search area of the filter. A variation on the Maximum Average Correlation Height (MACH) filter is used to create invariance to orientation while giving high 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 the regions of movement in successive video frames. Areas of movement producing a strong correlation response indicate an in-class target and can then be used to determine the position, in-plane rotation and scale of the target objects in the scene and track it over successive frames.

Distortion-tolerant correlation filter methods have been applied to many video-based automatic target recognition (ATR) applications, but in a single-frame architecture. In this paper we introduce an efficient framework for combining information from multiple correlation outputs in a probabilistic way. Our framework is capable of handling scenes with an unknown number of targets at unknown positions. The main algorithm in our framework uses a probabilistic mapping of the correlation outputs and takes advantage of a position-independent target motion model in order to efficiently compute posterior target location probabilities. An important feature of the framework is the ability to incorporate any existing correlation filter design, thus facilitating the construction of a distortion-tolerant multi-frame ATR. In our simulations, we incorporate the minimum average correlation energy Mellin radial harmonic (MACE-MRH) correlation filter design, which allows the user to specify the desired scale response of the filter. We test our algorithm on real and synthesized infrared (IR) video sequences that exhibit various degrees of target scale distortion. Our simulation results show that the multi-frame algorithm significantly improves the recognition performance of a MACE-MRH filter while requiring only a marginal increase in computation. We also show that, for an equivalent amount of added computation, using larger filter banks instead of multi-frame information is unable to provide a comparable performance increase.

In practical biometric verification applications, we expect to observe a large variability of biometric data and single classifiers may not be very accurate. In such cases, fusion of multiple classifiers may improve accuracy. Statistical dependence of classifiers has recently been shown to improve accuracy over statistically independent classifiers. In this paper, we focus on the verification application and theoretically analyze the OR fusion rule to find the favorable and unfavorable conditional dependence between classifiers. Favorably dependent correlation filter based classifiers for the OR rule are designed on the fingerprint NIST 24 plastic distortion and rotation datasets. For the plastic distortion dataset, unconstrained optimal tradeoff (UOTF) correlation filters were used because of their distortion tolerance and discrimination capability; and for the rotation dataset, optimal trade-off circular harmonic function (OTCHF) filters were used because of their tolerance to geometric rotation. On the plastic distortion dataset, three favorably dependent classifiers were designed on different distortions of the finger, each with an EER of 15.7%, 14.3%, and 9.8% respectively. The OR fusion of these three classifiers has an Equal Error Rate (EER) of 1.8% while the best single UOTF based classifier has an EER of 2.8%. On the rotation dataset, five OTCHF filter based classifiers were designed for tolerance to different rotation angle ranges of a finger with an average individual EER of 38.8%. The OR rule fusion has an EER of 14.6%; whereas the best single OTCHF filter has an EER of 27.7%. It is also shown that the best fusion rule is the OR rule for these classifiers that were designed to be favorable for the OR rule.

Various correlation based techniques for detection and classification of targets in forward looking infrared (FLIR) imagery have been developed in last two decades. Correlation filters are attractive for automatic target recognition (ATR) because of their distortion tolerance and shift invariance capabilities. The extended maximum average correlation height (EMACH) filter was developed to detect a target with low false alarm rate while providing good distortion tolerance using a trade off parameter (beta). By decomposing the EMACH filter using the eigen-analysis, another generalized filter, called the eigen-EMACH (EEMACH) filter was developed. The EEMACH filter exhibits consistent performance over a wide range which controls the trade-off between distortion tolerance and clutter rejection. In this paper, a new technique is proposed to combine the EEMACH and polynomial distance classifier correlation filter (PDCCF) for detecting and tracking both single and multiple targets in real life FLIR sequences. At first, EEMACH filter was used to select regions of interest (ROI) from input images and then PDCCF is applied to identify targets using thresholds and distance measures. Both the EEMACH and PDCCF filters are trained with different sizes and orientations corresponding to the target to be detected. This method provides improved clutter rejection capability by exploiting the eigen vectors of the desired class. Both single and multiple targets were identified in each frame by independently using EEMACH-PDCCF algorithm to avoid target disappearance problems under complicated scenarios.

Diffraction image correlator based on commercial digital SLR photo camera is described. The correlator is proposed for recognition of external 2-D and 3-D scenes illuminated by quasimonochromatic spatially incoherent light. Principal optical scheme of the correlator is analogous to that of incoherent holographic correlator by Lohmann. The correlator hardware consists of digital camera with attached optical correlation filter unit and control computer. No modifications have been introduced in units of commercial digital SLR photo camera. Digital camera was connected via camera interface to computer for controlled camera shooting, transfer of detected correlation signals and post-processing. Two ways were used for correlation filter unit mounting. In the first case, correlation filter was attached to the front of the camera lens. In the second one, it was placed in a free space of the SLR camera body between the interchangeable camera lens and the swing mirror. Computer generated Fourier holograms and kinoforms were used as correlation filters in experiments. The experimental setup of the correlator and experimental results on images recognition are presented. The recognition of test objects of direct and reversed contrast with the same correlation filter was performed.

Synthetic aperture radar (SAR) automatic target recognition (ATR) based on the extended maximum average correlation height (EMACH) distortion invariant filter (DIF) is presented. Prior work on the EMACH filter addresses 3-class and 10 class classification with clutter rejection. However, the ability of the EMACH filter to reject confusers is not well known. This paper addresses this. We follow a benchmark procedure which involves classification of three object classes over 360° aspect angle differences and with depression angle and variant differences and rejection of two unseen confusers from the Moving and Stationary Target Acquisition and Recognition (MSTAR) public database. We present a scheme to select which training set images to include while making the filters, since it is not necessary to use all training set images to make the filters. Results for classification with both confuser and clutter rejection are presented. We also compare our work with prior EMACH MSTAR work. We find EMACH filters to have poor confuser and clutter rejection. We also correct prior EMACH clutter rejection performance results.

We propose an image search engine that integrates collinear holography and the optical correlation technology used in FARCO. From preliminary correlation experiments using the collinear optical setup, we achieved excellent performance of high correlation peaks and low error rates. We expect optical correlation of 10 μs/frame assuming 12,000 pages of hologram in one track rotating at 600rpm. It follows that it is possible to take correlation at the speed of more than 100,000 faces/s when applied to face recognition. This system can also be applied for High-Vision image searching.

Color pattern recognition techniques involve the separation of basic color components, red, green and blue, by using color filters. Although several joint transform correlation architectures have been proposed in literature for color pattern recognition, however, these algorithms are suitable for single color target detection only and most of them are sensitive to noise and do not efficiently utilizes the space bandwidth product. A new shifted phase-encoded fringe-adjusted joint transform correlation (SPJTC) technique has been proposed in this paper for class-associative color pattern recognition. The color images are first split into three fundamental color components and the individual components are then processed simultaneously through three different channels. The SPJTC technique for each color component again involves two channels, one with the reference image and the other with 180° phase-shifted reference image. Both are phase masked using a random phase and then used with the input scene. The joint power spectra (JPS) are again phase masked and subtracted one from the other. The resultant JPS yields the desired correlation after inverse Fourier transformation. A modified class-associative color fringe adjusted filter is developed for providing single and sharp correlation peak per target while satisfying the equal correlation peak criterion for each class member. The salient feature of the proposed scheme is that the number of channels and processing steps remains constant irrespective of the number of members in the class. Computer simulation verifies the effectiveness of the proposed technique for color images both in binary and gray levels even in presence of noise.

We present a novel synthetic discriminant function (SDF) formulated from the Laplacian-enhanced (L) training images for the rotation and scale invariant target detection. It is shown that the proposed LSDF yields significantly improved correlation performance compared to the traditional SDF. Since the LSDF is formulated off line, it does not have any burden on the processing speed of the system.

Established optical correlation techniques will be applied to the problem of pose estimation of spacecraft for autonomous rendezvous and docking. The problem is substantially similar to the recognition and tracking of hostile targets in military applications. The historically deep knowledge base in this area will be exploited for the current investigation so that important discoveries will not have to be reinvented. It is expected that this problem should be somewhat better behaved since the "target" is cooperative and not actively in opposition to being found or recognized. The goal of this investigation is to determine the parameters of interest and to construct somewhat meaningful demonstrations of techniques that could be used for this optical approach to a computationally intense digital problem.

We have developed a mapping algorithm for correcting sinusoidally scanned images from their distortions. Our algorithm is based on an approximate relationship between linear and sinusoidal scanning. Straightforward implementation of this algorithm showed that the mapped image has either missing lines or redundant lines. The missing lines were filled by fusing the mapped image with its median filtered version. The implementation of this algorithm shows that it is possible to retrieve up to 96.43% of the original image, as measured by the recovered energy.

High energy physics, climate computations, nanoscience, fusion energy, astrophysics and genomics are applications with high processing and network demands. Optical components can be useful for these application as they can provide ultra fast, high input/output processing and network switching parts. In this paper a core concept is presented that may allow the systematic programming of linear optical components for optoelectronic processors, network switching or have general digital functionality. In this paper we are dealing with with a fundamental optical digital design concept. An optical automated logic design process is described, under a linear optics model assumption. We use optimization theory and maximum feasibility set (MAX-FS) inspired heuristics to solve the problem of finding optimal performance weights and optical thresholds for the implementation of a digital/switching function with linear optics. This optical design automation (ODA) may evolve into a rapid prototyping environment for fabless opto-electronics companies to receive custom programming for opto-electronic circuits from system engineers. Using this process, we have successfully high-level designed an 8-bit function using a single optical stage and a minimal electronic component.

A three dimensional (3-D) Multidetector Computed Tomography (MDCT) liver volume dataset of a patient with colorectal cancer was employed, each slice containing 512 x 512 pixels with pixel dimensions of 0.84 mm x 0.84 mm and slice reconstruction at a 2 mm interval. The 2-D slices were down-sampled to obtain a pixel width of 2mm x 2mm, thus resulting in an isometric cubic voxel of dimension of 2mm³. A 3-D Laplacian of Gaussian filter mask was generated for different pass-band regions in order to highlight fine to coarse texture within the volume. 3-D Fourier transformations of the liver volume and filter mask were obtained and their product computed. The filtered volume in the spatial domain was obtained by taking the inverse 3-D Fourier transform of this product. This 3D filtered volume highlighted textural features within different bands of spatial frequencies. The 3-D texture of the liver was viewed using maximal intensity projection. We quantify the texture over different spatial scale ranges for 3-D filtered liver volume which may have the potential of improving our preliminary results of 2-D texture analysis of liver images of patients with colorectal cancer and therefore could be employed as a computer assisted diagnostic tool in medical cancer imaging.

The paper presents an advanced 3D imaging system based on a combination of stereo vision and light projection methods. A single digital camera is used to take only one shot of the object and reconstruct the 3D model of an object. The stereo vision is achieved by employing a prism and mirror setup to split the views and combine them side by side in the camera. The advantage of this setup is its simple system architecture, easy synchronization, fast 3D imaging speed and high accuracy. The 3D imaging algorithms and potential applications are discussed. For ATR applications, it is critically important to extract maximum information for the potential targets and to separate the targets from the background and clutter noise. The added dimension of a 3D model provides additional features of surface profile, range information of the target. It is capable of removing the false shadow from camouflage and reveal the 3D profile of the object. It also provides arbitrary viewing angles and distances for training the filter bank for invariant ATR. The system architecture can be scaled to take large objects and to perform area 3D modeling onboard a UAV.

As we published last year, we have developed a very efficient image pre-processing scheme for using in any image analyzing system or any pattern recognition system. This scheme will analyze an edge-detected binary image and break it down to many simple branches by the automatic line tracking method we published earlier. Each branch can then be curve-fitted with the standard Window operations and it will result in an analog output which contains the starting point xy-coordinates, the ending point xy-coordinates, the polynomial degree, the coefficients in the best-fit algebra expression, and the angle of rotation to make the polynomial fitting work. The original binary image then can be closely reconstructed using this compact analog data. The reconstructed image is seen to be highly accurate compared to the original image in all our experiments. This paper reports the description of the topological structure of the original binary image detected by this novel image pre-processing method. That is, it will tell us how many branching points, how many single-ended points will be detected, and what algebraic curves are connected among them. This "topological" description of the image is not only very specific, but also very robust because when the image is viewed in different elevations and different directions, even though the geometrical shape changes, the topological or syntactical description will NOT change. Therefore it can be used in very fast learning, and very robust, yet very accurate, real-time recognition.

Remote sensing data have been widely used for land cover mapping using supervised and unsupervised methods. The produced land cover maps are useful for various applications. This paper examines the use of remote sensing data for land cover mapping over Saudi Arabia. Three supervised classification techniques Maximum Likelihood, ML, Minimum Distance-to-Mean, MDM, and Parallelepiped, P were applied to the imageries to extract the thematic information from the acquired scene by using PCI Geomatica software. Training sites were selected within each scene. This study shows that the ML classifier was the best classifier and produced superior results and achieved a high degree of accuracy. The preliminary analysis gave promising results of land cover mapping over Saudi Arabia by using Landsat TM imageries.

In the article it is presented a PC based system for persons' identification and verification on the basis of face correlation recognition. There are described the structure and the functions, the software, the interfaces, the options of image processing. There are presented the investigation results of the influence of the noise, rotation and scale of the input images on the identification process. There are calculated the data concerning the recognition time for the images of different resolution. In order to increase the system's productivity it is proposed to use an optical-electronic system.

Simple grating filters such as rectangular and triangular ones are used in a two non-conventional real-time joint-transform correlation (JTC) architectures to eliminate the spatial light modulators (SLM) at the Fourier planes of a conventional JTC. In the first one, the grating filter is used along with a heterodyning and one-dimensional optical scanning techniques to capture the cross-correlation functions of the input images without major processing. This significantly reduces the time processing needed for real-time applications by eliminating the drawbacks of the non-ideal characteristics of the SLMs. In the second technique, the one-dimensional optical scanning is eliminated to achieve a faster (but a little bit more complicated) processing.