This PDF file contains the editorial “Is referencing an art or an obligation?” for OE Vol. 39 Issue 05

This PDF file contains the editorial “Guest Editorial: Special Section on Target Recognition Techniques” for OE Vol. 39 Issue 05

Several filter algorithms are developed for the purpose of modifying the transmissivity of many matched filters multiplexed onto one hologram. Each algorithm optimizes, or approximately optimizes, a specified performance metric. Four pages of images are multiplexed onto a thin hologram, each page having nine binary images. Computer simulation indicates that some algorithms operate over a wide range of obscurations (down to about 5% of the original image), while others have restricted operating ranges. This process has application to both associative memories and correlation. Two new optical schemes are also proposed that can implement two of these modification filters.

we show that the maximum average correlation height (MACH) correlation filter overemphasizes the importance of the average training image leading to poor discrimination of the desired class images from clutter images. To overcome this, two new metrics termed the all image correlation height (AICH) and the modified average similarity measure (MASM) are introduced and optimized in a new correlation design. The resulting filter exhibits improved clutter rejection performance while retaining the attractive distortion tolerance feature of the MACH filter. Simulation results based on a simulated synthetic aperture radar (SAR) image database are presented to illustrate the new filter's properties.

We introduce a new fixed-phase algorithm for producing an object recognition correlation filter by either enhancing or suppressing unnecessary spectrum information from selected objects. We demonstrate this approach by simulating the performance using five test images with completely dissimilar edge information. This new algorithm can be used to substantially improve the performance of existing composite classification and recognition filter algorithms.

An amplitude-modulated radial harmonic filter is introduced to improve the discrimination of the radial harmonic filter for shift- and scale-invariant pattern recognition. The improvement of discrimination is achieved by Fourier-domain amplitude modulation, which places more emphasis on the high-frequency terms. The gain of better discrimination, however, is at the cost of poorer noise robustness since the amplitude- modulated radial harmonic filter is an optimal trade-off filter for noise robustness and discrimination. To compensate for the loss of noise robustness, we apply the technique of Fourier-domain amplitude truncation to the proposed filter. With a given size of bandpass, the amplitude truncation approach results in the lowest signal energy loss, thus yielding a better SNR and discrimination than with a low-pass amplitude-modulated radial harmonic filter.

The ring synthetic discriminant function (RSDF) filter for rotation-invariant response is discussed for pattern recognition. This method uses one half of a slice of the Fourier transform of the object to generate the transfer function of the filter. This is accomplished by rotating the one half of a slice in the Fourier domain through 2? rad about the zero-frequency point of the Fourier plane. This filter has the advantage of always matching at least one half of a slice of the Fourier transform of any rotation of the image. An analytical discussion of the filter construction and correlation results are presented along with simulated correlation results for a particular target image. These results and established metrics are used for comparison with benchmark algorithms.

We use circular versions of advanced distortion-invariant filters such as optimal trade-off synthetic discriminant function and distance classifier correlation filters to obtain rotation invariance with an optical correlator. The filter noise performance is compared using a common measure of probability of error because the filters have different characteristics. The filters are real-valued so they can be implemented on a variety of SLMs. The circular symmetry of the filters significantly decreases their computational requirement.

A novel phase-encoded fringe-adjusted joint transform correlator (JTC) (pFJTC) technique is proposed that employs phase encoding and nonlinear Fourier plane apodization to optimize correlation performance. For input scenes involving multiple identical targets, existing JTCs require multistep on-line processing to eliminate the false alarms and other artifacts. The proposed pFJTC eliminates the possibility of false target detection by eradicating the false correlation peaks and alleviates the effects of noise and other artifacts; in just one step, thus ensuring higher processing speed. The phase-encoding operation and the Fourier plane apodization have no detrimental effect on the processing speed. This technique ensures better utilization of the space- bandwidth product by using separate input and reference planes and by yielding only one peak per target instead of a pair of peaks produced by alternative JTCs. An all-optical implementation for the pFJTC technique is proposed and computer simulation results are presented.

We perform computer simulations and theoretical modeling to predict the location of the best operating point of the two-beam coupling joint transform correlator. From this study, we determine that the highest speed and resolution as well as the optimal trade-off between correlation peak intensity, efficiency, and noise performance is realized when the holographic material is thin and operates with relatively small negative gain.

We propose the implementation of normalized correlation in the space domain to achieve real-time discrimination. Two new joint transform correlator configurations are proposed for this purpose. Compared to the filter normalization method, the proposed schemes are more suitable for real-time processing and are robust in terms of system alignment. Computer simulation results are given.

We propose and demonstrate new wavelet transforms, the modified Haar wavelet transform and Haar-Gaussian wavelet transform, for the detection of central positions of a noisy rounded square function. The advantage is the immunity of the wavelet transform to random noise. The alternative technique based on first- and second-order differentiation cannot be applied when the signal function is corrupted by noise. The proposed technique can be used for lead inspection of surface mount devices in the electronic industry.

A forest representation of 2-D wavelet transform is proposed to emphasize a local signal and coefficient propagation across multiscales. A multiresolution analysis involves a forest consisting of an approximate forest and three detail (horizontal, vertical, and diagonal) forests. An approximate tree is a collection of successive approximations at multiscales at a location, and a detail tree is a collection of detail coefficients along an orientation. In fine-to-coarse processing, local neighboring trees merge into a bulky tree recursively and thereby incur large uncertainty in signal location. The coefficient propagation across multiscales can be used for local signal detection.

An improved multiple target detection technique is proposed by using a wavelet joint transform correlator (JTC). A bank of wavelet filters are generated using the Mexican hat wavelet function and are 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 for both binary and gray-level input scenes while avoiding the false alarms as well as rejecting the nontarget objects. An all-optical implementation for the proposed wavelet JTC technique is also suggested. Computer simulation results are presented to confirm the effectiveness of the proposed technique.

We propose a novel orthogonal wavelet-based approach to reduce speckle in synthetic aperture radar and gray-scale images for better automatic target recognition applications. We also analyze different image metrics using different thresholding and image data compression techniques.

The projection-slice synthetic discriminant function, which was described earlier, reduces significant portions of the training data in the synthesis process as compared to the normal synthetic discriminant function filters. This enables the matching of particular features of each training image. The wavelet transform is utilized to further reduce the data in the synthesis stage and thereby reduce the data overlap in filters. The method for this type of filter synthesis is discussed and simulation results of the algorithms involved are presented to show that the relative performance of the newly synthesized filter is improved, while the data representing the training images has been reduced.

The ability to accurately predict natural backgrounds over a wide range of potential image acquisition conditions is key to wide-area change detection. The lack of robustness in pure image-based techniques has led to the development of site modeling and model-supported exploitation approaches. Instead of using multiple images to build a site model, we consider the possibility of using them directly to estimate the background of a new image for change detection. Our method uses a nonlinear mean-squared estimation technique to compute image backgrounds from multiple reference images. In comparing linear and nonlinear estimators, we find that although the performance of both improves as the number of reference images increases, the nonlinear estimator yields significantly better background estimates and is much less sensitive to registration errors between images than the linear estimator. We show that as the number of references increases, the pixels within a region become physically significant; i.e., they correspond to parts of the scene with similar physical properties. Synthetic aperture radar (SAR) background modeling and change detection results including a crosssensor change detection example in which SAR reference images are used to predict the background of an electro-optical (EO) image are presented. Significant performance gains over linear change detection (up to two orders of magnitude reduction in the false alarm rate) are demonstrated in each case over a 60-deg range of SAR sensor aspect angles.

We describe a new architecture for synthetic aperture radar (SAR) automatic target recognition (ATR) based on the premise that the pose of the target is estimated within a high degree of precision. The advantage of our classifier design is that the input space complexity is decreased with the pose information, which enables fewer features to classify targets with a higher degree of accuracy. Moreover, the training of the classifier can be done discriminantly, which also improves performance and decreases the complexity of the classifier. Three strategies of learning and representation to build the pattern space and discriminant functions are compared: Vapnik's support vector machine (SVM), a newly developed quadratic mutual information (QMI) cost function for neural networks, and a principal component analysis extended recently with multiresolution (PCA-M). Experimental results obtained in the MSTAR database show that the performance of our classifiers is better than that of standard template matching in the same dataset. We also rate the quality of the classifiers for detection using confusers, and show significant improvement in rejection.

Target recognition can greatly enhance the usefulness of radar systems. For speed, optical architectures are the preferred method of implementation for the classifiers. Both correlators and feature-based neural networks have elegant optical implementations. The choice of which recognition system to use is problem-dependent. We evaluate the relative performance of correlation- and feature-based (neural network) classifiers on a set of four simulated radar targets over a wide range of target orientations. Experiments are performed for a range of radar bandwidths to determine the effect of radar bandwidth on the relative classification performance. Only 1-D radar range profiles are considered since it is assumed that the targets are classified using few profiles and the orientation of the target in each profile is known only approximately. The results suggest that feature-based classifiers outperform correlationbased classifiers and that classification performance is highly dependent on the orientation of the aircraft, but that accurate classification of approaching targets is possible

The design of composite filters for illumination-invariant face recognition is discussed. As 3-D profiles have been extensively used as a way to ensure this behavior, the two main constraints of 3-D analysis (i.e., 3-D profile acquisition and the 3-D face-matching computational requirements) are considered. The first limitation is solved using a 3-D acquisition system based on the projection of structured color light, which provides reliable 3-D data with a very simple setup. The second shortcoming is tackled by limiting the recognition task to 2-D complexity, so the computational cost is reduced. With the 3-D profile information available, 2-D training set images that correspond to different illumination conditions are derived for the synthesis of a 2-D composite filter using the simulated annealing algorithm. The experimental results confirm the potentiality of the proposed method. Additionally, the concept of ''average face" is presented and its feasibility as an appropriate method for the recognition of a target from a set of objects is supported by experimental results.

we show that the Fourier-Mellin moments can provide an invariant and orthogonal description for binary and contour images. we use the cubic spline interpolation of the feature space trajectories to improve the linear interpolation for 3-D object recognition from 2-D perspective views.

. A typical target recognition system usually involves three stages: image enhancement, image segmentation, and object identification. We propose an efficient image segmentation approach to segment man-made targets from unmanned aerial vehicle (UAV) imagery using curvature information derived from image histogram smoothed by Bezier splines. The experimental results show that by enhancing the histogram instead of the original image, similar segmentation results can be obtained in a more efficient way. Algorithm complexity is analyzed. Segmentation results from images enhanced by variants of diffusion methods are also presented for comparison. Real-time automatic target recognition (aTr) is the main objective of the proposed approach.

Subpixel detection in multispectral imagery presents a challenging problem due to relatively low spatial and spectral resolution. We present a generalized constrained energy minimization (GCEM) approach to detecting targets in multispectral imagery at subpixel level. GCEM is a hybrid technique that combines a constrained energy minimization (CEM) method developed for hyperspectral image classification with a dimensionality expansion (DE) approach resulting from a generalized orthogonal subspace projection (GOSP) developed for multispectral image classification. DE enables us to generate additional bands from original multispectral images nonlinearly so that CEM can be used for subpixel detection to extract targets embedded in multispectral images. CEM has been successfully applied to hyperspectral target detection and image classification. Its applicability to multispectral imagery is yet to be investigated. A potential limitation of CEM on multispectral imagery is the effectiveness of interference elimination due to the lack of sufficient dimensionality. DE is introduced to mitigate this problem by expanding the original data dimensionality. Experiments show that the proposed GCEM detects targets more effectively than GOSP and CEM without dimensionality expansion.

The recognition of targets in IR scenes is complicated by the wide variety of appearances associated with different thermodynamic states. We represent variability in the thermal signatures of targets via an expansion in terms of "eigentanks" derived from a principal component analysis performed over the target's surface. Employing a Poisson sensor likelihood, or equivalently a likelihood based on Csiszar's I-divergence (a natural discrepancy measure for nonnegative images), yields a coupled set of nonlinear equations which must be solved to compute maximum a posteriori estimates of the thermodynamic expansion coefficients. We propose a weighted least-squares approximation to the Poisson loglikelihood for which the MAP estimates are solutions of linear equations. Bayesian model order estimation techniques are employed to choose the number of expansion coefficients; this prevents target models with numerous eigentanks in their representation from having an unfair advantage over simple target models. The Bayesian integral is approximated by Schwarz's application of Laplace's method of integration; this technique is closely related to Rissanen's minimum description length criteria. Our implementation of these techniques on Silicon Graphics computers exploits the flexible nature of their rendering engines. The implementation is illustrated in estimating the orientation of a tank and the optimum number of representative eigentanks for both simulated and real data.

The detection of venous beading in retinal images provides an early sign of diabetic retinopathy and plays an important role as a preprocessing step in diagnosing ocular diseases. We present a computer-aided diagnostic system to automatically detect venous beading of blood vessels. It comprises of two modules, referred to as the blood vessel extraction module (BVEM) and the venus beading detection module (VBDM). The former uses a bell-shaped Gaussian kernel with 12 azimuths to extract blood vessels while the latter applies a neural network-based shape cognitron to detect venous beading among the extracted blood vessels for diagnosis. Both modules are fully computer- automated. To evaluate the proposed system, 61 retinal images (32 beaded and 29 normal images) are used for performance evaluation.

Experimentally obtained spectral dependencies of an optical signal influenced by intermodal interference in a fiber under various conditions (for example, different lengths or different fibers) are presented. It is shown that the difference between the propagating constants of the first two modes can be obtained from the measured spectrum. The obtained results illustrate the usefulness of such investigation for determination of basic parameters of fibers, especially for investigating the homogeneity of a produced fiber.

The polarization-maintaining (PM) capability of a high- birefringence (hi-bi) optical fiber can be described in terms of its modal birefringence or beat length. In the hi-bi (PM) fiber the beat length can be measured by changing the state of polarization of the light through an external means. One such technique is based on the Faraday magnetooptic effect. The technique is nondestructive, relatively simple, and efficient, and beat lengths of 2 mm and less have been measured accurately with it, thus fulfilling the measurement requirements for many of the currently available PM fibers. Utilizing optical sources with wavelengths of 633 and 814 nm, the beat lengths in a total of four hi-bi fibers have been successfully measured using the dc Faraday magneto-optic effect. The experimental results are comprehensive and very promising, and it has been demonstrated that this nondestructive technique has significant advantages over many others in simplicity, versatility, productivity, practicality, and accuracy.

Results on scattering, off-center coupling, and other losses that influence the shape of a beam delivered by flexible hollow waveguides are presented. The conditions for the beam shape to be near to that of the source are found. A theoretical model for the radiation propagation through the hollow waveguides and the conditions for change in the beam shape are calculated. Good agreement between the theoretical and experimental results was obtained.

A new method for three-dimensional shape recovery from image focus is proposed. The method is based on approximation of the focused-image surface (FIS) by a piecewise curved surface that tracks the realistic FIS in image space. The piecewise curved surface is estimated by interpolation using the Lagrangian polynomial. The new method has been implemented on a prototype camera system. Experiments and their results are described and discussed. The results show that the new method is more accurate than previous methods.

It is well established that lattice quantizers are highly suitable for encoding wavelet coefficients of still images. Existing lattices give rise to the variety of lattice quantizers available for quantization. In this paper we address the problem of selecting a lattice quantizer for a wavelet- based image coding system and show that the rate-distortion performance of a lattice quantizer can be well estimated based on the generalized Gaussian model of wavelet coefficients of images. Simplification of the rate-distortion function used for the estimation is achieved by utilization of the theta series with the 0.7-norm, which are derived in this paper. As shown by experimental results, our estimation closely matches the real rate-distortion performance of lattice quantizers.

Vector quantization (VQ) has been accepted as one of the most effective image compression methods with provable rate-distortion optimality. The outputs of VQ are a collection of indices, which correspond to the addresses of the codevectors in the codebook. The indices are, however, not mutually independent. They are in fact very highly correlated and are thus appropriately described by a Markov system. In this paper, a Markov system for VQ indices is introduced. Statistics are gathered for various scans, such as the zig-zag, Peano, row-major and column-major scans. The proposed method, like address VQ, achieves the same image quality as conventional VQ. Simulation results show that the proposed method achieves a better bit-rate reduction than Address- VQ. Besides, both the computational complexity and memory needed for the proposed method are lower. Nevertheless, the only extra operation needed by the proposed method is a simple table retrieval operation on both the encoder side and the decoder side. We believe that it is a method worth further exploration.

Advance knowledge of the time required by an observer to detect a target visually is of interest, e.g., in preparing flight scenarios, in modeling mission performance, in evaluating camouflage effectiveness, and in visual-scene generator calibration. A wide range of computational models has therefore been developed to predict human visual search and detection performance. This study is performed to test the quality of the predictions of three of these models: ORACLE, Visdet, and a formula by Travnikova. The three different models are used to predict the results of an experiment in which observers searched for military vehicles in complex rural scenes. The models predict either the mean time required to find the target, or the probability of finding the target after a given amount of time, from a few physical parameters describing the scene (the mean scene luminance, the angular dimensions of the field of view and the target, the intrinsic target contrast, etc.). None of the models reliably predicts observer performance for most of the scenes used in this study. ORACLE and Visdet both overestimate the detection probability for most situations. The formula by Travnikova does not apply to the scenes used here.

An adaptive Wiener filtering approach, based on image estimation in the wavelet domain, is proposed. Gradient-based estimation of the image is employed by minimizing an error functional that depends on estimates of the image and the power of the noise in each wavelet subband. The power of the noise is estimated from the variance of wavelet coefficients. The Wiener filter used in restoring the corrupted image is updated at each iteration step. Hard wavelet thresholding allows for appropriate selection of initial estimates of the restored image and the noise. The initial estimate of the restored image is then improved by updating these estimates iteratively. Conditions for convergence of the proposed procedure are derived. Experimental results for restoring images corrupted by colored noise are presented as well. Comparisons with conventional restoration techniques strongly favor the proposed method.

Recently, a method was developed for determining interface profiles of extreme-ultraviolet layered synthetic microstructures (LSMs), which is based on computer processing of digitized LSM electron micrographs. Nickel/carbon and molybdenum/carbon multilayers are studied and analyzed in order to show the potentialities of the method in controlling the uniformity of the period and the thickness of the layers when LSMs are manufactured. Such a method makes it possible to get information about the quality of LSMs on a nanometer scale.

A new automatic target detection method for IR images that only requires information about the size of the targets is described. The proposed nonlinear-enhancement-based system (NLEBS) algorithm is based on a nonlinear enhancement paradigm that increases the contrast of the targets with minimal change in the clutter's contrast. The NLEBS employs several stages of processing, each with a different operational purpose. First, the nonlinear enhancement operation is performed by using an iterative procedure. After binarization, segmentation merging causes each local image region to grow by filling in holes. Then segmentation pruning is applied to remove spurious segments. Finally, a heuristic-based metric is employed to validate the possible targets. The performance of the NLEBS was tested with a large set of IR images. The results of these experiments showed a probability of detection greater than 90% and a false-alarm rate of about 1 false alarm per image.

By inserting an As2S3 optical fiber directly into a coupling hole of one of the optical resonator mirrors, a pigtail-type CO laser has been created. Its fundamental operation characteristics are described, together with theoretical treatments. First, by finding the spatial intensity distribution on the coupling-hole mirror, its effective transmittance t is evaluated in comparison with the measured spatial beam profile. Then, the output power is measured with various sizes of coupling-hole mirrors, obtaining the maximum output power at a coupling-hole diameter of ?0.6 mm. The dependence of the output power on the coupling-hole diameter (or the effective transmittance t) is theoretically explained by using the gain parameter G, and the internal loss ã, is obtained in good agreement with the experiments. It follows that the optimum coupling- hole diameter as evaluated is quite suitable for such a laser. Possible applications of this scheme to computer-aided robotics and automatic machinery are also described.

We have designed and constructed a CO2 laser marking system based on a scanned flexible waveguide (silver halide infrared optical fiber or a hollow glass waveguide). CO2 laser power was introduced into the proximal end of the waveguide and the distal tip was placed near the surface of a sample. A permanent magnet was attached to the distal tip, which was displaced in the X and Y directions, using a varying electro magnetic field. An on-line computerized system was used to control the laser power, the position of the distal tip and the scanning speed. The system is capable of cutting, heating or marking on various surfaces. The output profile distribution, spot size and resolution of hollow waveguides and silver halide fibers were investigated and compared.

The benchmark problem addresses the efficient allocation of an agile-beam radar in the presence of highly maneuverable targets and radar ECM. The multisensor benchmark tracking solution is aided by the presence of a scanning infrared search-and-track (IRST) system. This paper presents methods for applying a tracker with interacting multiple- model (IMM) filtering and multiple-hypothesis tracking (MHT) data association to this multiple-sensor tracking and resource allocation problem. It presents a hybrid multisensor tracking architecture in which an IMM- MHT tracker operating on IRST data provides the global IMM-MHT tracker with selected observations. Simulation results quantify the potential improvement from the use of advanced tracking methods and IRST data to enhance agile-beam radar tracker capability.

An analysis is performed of the degradation of the signal readout of fiber optic Bragg grating sensors demodulated using the serrodyne interferometric technique. Experimental results are presented that validate the analysis. It is also shown that synchronous detection of the interferometric signal referenced to the serrodyne waveform that modulates the phase of the readout interferometer allows the elimination of the system dc offset introduced by the serrodyne flyback.

Design and packaging of simple, low-cost, noncontact optical sensors for continuous liquid-level measurements are discussed. The sensors considered consist of only a source and a photoreceptor placed in a cylindrical tube containing the liquid. Selection of the sizes of various devices and placement of these devices in a compact package while maintaining suitable distances between them are considered. Various parameters of the possible packages are selected through diffraction calculations.

A precise condition on the spatial resolution of the IR collimator for testing thermal imaging systems is presented. The condition can be used even if only the spatial resolution of the IR collimator and that of the system under test, measured using popular definitions, are known. It is shown that when the condition is fulfilled, the thermal image degradation caused by the IR collimator is negligible.