Several approximations lead to a feasible method for calculating the effect of multi-path radar scattering at the sea surface. ONe of these is radial symmetry of the scattered microwave energy about the specular direction, according to which the scattered intensity depends on the 'Specular Deviation Angle'. This angle exhibits a stochastic variation that depends on the statistics of the sea surface and determines the average intensity of the scattered microwaves.

Gaussian receivers perform poorly in detecting small targets in non-Gaussian clutter, but significant improvement is available by using an appropriate non-Gaussian receiver. Selection of the proper receiver requires an adequate characterization of the unknown probability density function (PDF) of the clutter. In applications where the clutter environment changes and a limited number of homogeneous samples are available to approximate the PDF, an efficient algorithm is essential. The Oeztuerk PDF approximation algorithm satisfies this requirement, needing on the order of 100 samples from the PDF to be approximated.

This paper presents a comparison of chaotic and statistical CFAR detectors for detection of manmade point targets from SAR. Detection of small manmade targets in SAR or IR clutter is an important area of interest in many applications such as ocean surveillance, search and rescue, remote sensing, mine detection, etc. It has been shown that IR and radar clutter exhibit chaotic rather than purely random behavior. From the chaotic point of view, a neural network predictor has been developed using Radial Basis Functions (RBF) to detect small targets embedded in natural clutter. In this paper, we present tradeoff studies between the above chaotic CFAR detector and purely statistical detectors such as the Cell Averaging, Order Statistics, and Optimal Weibull. The tradeoff studies are performed on real data with real or simulated targets. It is shown that adaptive chaotic RBF detectors many outperform statistical detectors in real clutter environments.

Many remote sensing applications require comparison between satellite images acquired over a span of years. Commercially available radar images, such as ERS images, have misalignments between images ranging up to 100 pixels for images acquired a few years apart. The registration of these images is rendered difficult due to three reasons. (i) The volume of data to be processed makes speed of execution key issue. (ii) The uncertainty regarding scene contents rules out the use of any predefined 'target class'. (iii) The long time interval between images leads to large scene content variations. This requires the matching process to be very robust.

This paper addresses the joint estimation of backscatter and extinction coefficients from range/time noisy data under a nonlinear stochastic filtering setup. This problem is representative of many remote sensing applications such as weather radar and elastic-backscatter lidar. A Bayesian perspective is adopted. Thus, in addition to the observation mechanism, relating in a probabilistic sense the observed data with the parameters to be estimated, a prior probability density function has to be specified. We adopt as prior a causal first order auto-regressive Gauss-Markov random field. By using a reduced order state-space representation of the prior, we derive a nonlinear stochastic filter that recursively computes the backscatter and extinction coefficients at each site. A set of experiments based on simulated data illustrates the potential of the proposed approach.

This article is devoted to the experimental study of the generalized detector and comparison with the optimal detectors of classical and modern signal detection theories. Experimental investigations are carried out under the power signal-to-noise ratio (SNR) equal to 15.92 dB and 0.96 dB at the inputs of detectors. The signal is clearly detected by the generalized detector. But 0.96 dB is the region of the failure to detect signals by the optimal detectors. New features of signal detection, determination and estimation of the signal parameters using the generalized detector are discussed. Main functioning principles of the generalized detector are discussed. The purpose of the correlation and autocorrelation channels of the generalized detector is defined. The practical recommendations for employment of the generalized detector in various complex signal processing systems are discussed.

We proposed to use the possibility of recognition of the targets on background of the scattering from the surface, weather objects with the help of polarimetric 3-cm radar. It has been investigated such polarization characteristics: the amplitudes of the polarization matrix elements; an anisotropy coefficient; depolarization coefficient; asymmetry coefficient; the energy section was less than 1 dB at ranges up to 15 km and less than 1.5 dB at ranges up to 100 km. During the experiments urban objects and 6 various ships of small displacement having the closest values of the backscattering cross-section were used. The analysis has shown: the factor of the polarization selection for anisotropy objects and weather objects had the values about 0.02-0.08 Isotropy had the values of polarimetric correlation factor for hydrometers about 0.7-0.8, for earth surface about 0.8-0.9, for sea surface - from 0.33 to 0.7. The results of the work of recognition algorithm of a class 'concrete objects', and 'metal objects' are submitted as example in the paper. The result of experiments have shown that the probability of correct recognition of the identified objects was in the limits from 0.93 to 0.97.

Conditions have been identified from interceptor seeker data analysis that may degrade the performance of background estimation and object detection algorithms. A study of various algorithms was conducted to test the robustness of these algorithms in a variety of scenarios on real flight test interceptor seeker data. The goal of the study was to find one algorithm that would perform well with limited computer resources regardless of the scenario. This paper will describe each algorithm tested, discuss the benefits and issues of each technique, and present the results and conclusions.

Wavelet-based detection algorithms are developed for detection small targets in non-Gaussian clutter. A wavelet transform is applied to reduce spatial correlation of the clutter. An adaptive matched filter is applied in the wavelet transform domain which uses estimated covariance matrices derived from the wavelet coefficients. Two problems hinder the use of covariance estimates for background clutter removal: slow computational speed and induced false alarms resulting from nearly singular covariance estimates due to small sample sizes. The issue of speed is dealt with by evaluating the covariance matrices on a sparse grid followed by low order interpolation. To control the problem of bad covariance estimates we filter the grid generated covariance matrices to remove outliers using peer group averaging. This procedure removes the false alarm problem associated with nearly singular covariance estimates without degrading the overall performance of the clutter removal process.

A method of detecting dim targets in highly-cluttered time- varying image sequences is presented, where reliable clutter rejection is achieved by calibrating the multivariate statistics of a small number of generic space-time filters. The targets have sufficiently low SCR that a track-before- detect method is required. For targets where there is little prior information on velocity, a large number of filters is generally required to achieve a high response relative to the background. In the method described here, instead of applying thresholds to individual filters, joint filter statistics are used to estimate conditional threshold exceedance probabilities. A smaller number of more generic filters are applied, which are not finely tuned to targets but which characterize aspects of both targets and clutter. Potential targets are cued based on a non-parametric estimate of the probability of occurrence of similar clutter. Constant false alarm rates are inherent in the method. The method is demonstrated on examples of real forward-looking imagery of the sea surface, where glint is a source of strong clutter. Dim targets are distinguished form clutter by using the joint statistics of three variables: a constant-intensity blob filter, a filter tuned to sea glint flashes, and the vertical image coordinate.

Cruise missiles over land and sea cluttered background are serious threats to search and track systems. In general, these threats are stealth in both the IR and radio frequency bands. That is, their thermal IR signature and their radar cross section can be quite small. This paper discusses adaptive sequential detection methods which exploit 'track- before-detect' technology for detection glow-SNR targets in IR search and track (IRST) systems. Despite the fact that we focus on an IRST against cruise missiles over land and sea cluttered backgrounds, the results are applicable to other sensors and other kinds of targets.

The ability to detect and track dim unresolved targets in heavy clutter can be improved by the inclusion of the spectral dimension. Because of the great variation in targets, operating conditions and environments factors the spectral signature of the target is typically unknown. This paper present a fully adaptive matched filter and tracking paradigm which assumes no a priori information about the spectral signature of the target. It is shown that the full SCR gain can be realized in the absence of the spectral signature of the target. The ROC curve of the detector is used to show that performance loss due to the absence of spectral information is entirely due to an increase in the false alarm probability. This increase in P_FA adversely effects tracker performance. The SCR track feature is developed to mitigate these effects. Track features provide an information shunt around the detection threshold nonlinearity that would otherwise block the flow of useful information to the tracker.

There are several methods reported in the literature for detecting dim targets against slowly moving clutter. However, each method has its won advantages and disadvantages. The challenge lies in reducing the false alarm rate to an acceptable level. 'False alarm rate' defined in case of a significant size of the target in a frame may not be applicable to point-targets. This paper presents a new method for the detection of dim point-targets in the presence of the evolving clouds and heavy background clutter. Choosing a threshold for achieving constant false alarm rate is always a tricky problem. Too less a threshold may ensure detection of target pixels. But this will result in too may false targets, which limit the performance of the post-processor to trace out target paths. Too high a threshold result in fewer false alarms but the targets may also be missing out. Based on off-line studies, it has been found that a 'desirable condition' is required to limit the number of accumulated pixels not to exceed 8 percent of the total image size for post-processing. This paper present a method based on random and correlated noises which in turn selects an auto threshold that leads to the 'desired condition' for the post-processor. An effort has been made to derive an empirical formula based on random and correlated noises to obtain an auto threshold value that achieves the desirable condition. Then the incoming frames of data are then processed by a constant threshold and accumulated as total number of pixels. At the same time the track record of pixels along with frame numbers are recorded. The post-processor to filter out the false alarms uses this information. One advantage of this method is that there is no need to store all the frames to obtain the desired information. The algorithm has been tested with the available database and the results are very promising. It is assumed that most of the targets occupy a couple of pixels. Head-on moving and maneuvering targets are not considered.

The 3D matched filter proposed by Reed et al. and its generalizations provide a powerful processing technique for detecting moving low observable targets. This technique is a centerpiece of various track-before-detect (TBD) systems. However, the 3D matched filter was designed for constant velocity targets and its applicability to more complicated patterns of target dynamics is not obvious. In this paper the 3D matched filter and BAVF are extended to the case of switching multiple models of target dynamics. We demonstrate that the 3D matched filtering can be cast into a general framework of optimal spatio-temporal nonlinear filtering for hidden Markov models. A robust and computationally efficient Bayesian algorithm for detection and tracking of low observable agile targets in IR Search and Track (IRST) systems is presented. The proposed algorithm is fully sequential. It facilitates optimal fusion of sensor measurements and prior information regarding possible threats. The algorithm is implemented as a TBD subsystem for IRST, however the general methodology is equally applicable for other imaging sensors.

This paper provides a real world target and clutter model for evaluation of radar signal processing algorithms. The procedure is given for target and clutter data collection which is then followed by the equalization and superposition method. We show how the model allows one to vary the target signal to clutter noise ratio so that system performance may be assessed over a wide range of target amplitudes, i.e. detection probability versus target signal to noise ratio. Three candidate pre-track algorithms are evaluated and compared using this model as input in conjunction with an advanced tracker algorithm as a post processor. Data used for the model represents airborne traffic operating over the body of water bounded by North, Central, and South America. The processors relate to the deployment of Over the Horizon Radar for drug interdiction. All the components of this work, model as well as the processors, are in software.

The SPIRIT system is a spectrally agile IR imaging airborne camera, with the capability to select any of the multiple filters on a frame by frame basis. The implemented solution employs advanced, but proven, technology to meet the objectives, and achieved good spatial and thermal performance in all modes. Sophisticated electronic design has results in a flexible unit, which can respond to the changing requirements of the user. Initial SPIRIT flight trials were undertaken in summer 1998 with more scheduled to continue through 1999. The sensor was installed on to DERA's TIARA research platform, a modified Tornado F2. The flight trials to date have been conducted over a variety of scenarios, collecting spectral data in up to 12 bands, of other aircraft, tanks, and fixed targets. Further ground- based trials, with the sensor mounted on a pan and tilt tracking platform, have been performed on characterized targets and against further air targets. Data from these initial trials are currently being processed to assess whether sufficient spectral information is available to discriminate between target types at militarily significant ranges. Sample hyperspectral imagery form SPIRIT and some results are presented.

A new system for robotically assisted retinal surgery requires real-time signal processing of the reflectance signal from small targets on the retina. Laser photocoagulation is used extensively by ophthalmologists to treat retinal disorders such as diabetic retinopathy and retinal breaks. Currently, the procedure is performed manually and suffers from several drawbacks which a computer-assisted system could alleviate. Such a system is under development that will rapidly and safely place multiple therapeutic lesions at desired locations on the retina in a mater of seconds. This system provides real- time, motion-stabilized lesion placement for typical clinical irradiation times. A reflectance signal from a small target on the retina is used to derive high-speed tracking corrections to compensate for patient eye movement by adjusting the laser pointing angles. Another reflectance signal from a different small target on the retina is used to derive information to control the laser irradiation time which allows consistent lesion formation over any part of the retina. This paper describes the electro-optical system which dynamically measures the two reflectance signals, determines the appropriate reflectance parameters in real time, and controls laser pointing and irradiation time to meet the stated requirements.

A simple and practical approach to observability analysis of bearings-only target tracking is developed. The emphasis is on discrete time cases. The uniqueness of this approach is that the observability properties are derived directly from the concise relationships between target bearings and the observability matrix, instead of the much more complicated relationships between own-ship state and the observability matrix. Some observability criteria are obtained very naturally using such an approach. Its potential in such applications as own-ship maneuver optimization and tracking algorithm development is also very prospective.

Target detection and tracking is a subject of great value to the defence community. With the need to build better detection and tracking system comes the need to evaluate the performance of such systems. Performance evaluation not only aids the system user in selecting the appropriate system for the specific application, but can also aid the developer in building and improving systems. An earlier paper discussed using the Neyman-Pearson (NP) criterion for evaluating the performance of tracking systems. The NP criterion is especially appropriate for evaluating the performance of tracking systems where the culminating action is an event such as the queuing of another systems or the engagement of a target. In this paper, some of the issues that have arisen since the earlier paper are addressed. First, a review of the application of the NP criterion to tracker performance assessment is given including a statement on the statistical significance of the declaration threshold setting. Performance is evaluated by examining lists of track declarations generated by running the system under test on real scenes without targets and on real scenes with targets inserted at given target strengths. One of the difficulties with the NP method of tracker performance evaluation is the computation and data storage requirements for setting the track declaration threshold. The use of models for reducing these requirements is discussed.

SUppose that a multitarget tracker is used to track a dim target in heavy clutter. In the paper 'Characteristics of the Acquisition of a Dim Target in Clutter', Oliver Drummond has pointed out that, for such a tracker, target acquisition is a very different problem than target detection. In particular, target acquisition requires two Receiver Operating Characteristic (ROC) curves; whereas target detection requires only one. In past presentations at this and other conferences and in the book Mathematics of Data Fusion, we have introduced 'finite-set statistics' (FISST), a direct generalization of conventional single-sensor, single-target statistics to the multitarget realm. In this paper we show how FISST results in a unification of both detection and acquisition under a familiar decision- theoretic framework. The basic idea is to use a unified Bayesian single-target tracker with clutter models, rather than a multitarget tracker. In analogy to a conventional detection problem, the acquisition problem is reduced to a conventional decision problem. In particular, we show how to define an 'acquisition ROC curve' for this sort of generalized detection problem.

Multiple Target Tracking (MTT) requires the proper association of measurement data with individual targets before the data is filtered and target positions estimated. Because data association is computationally intensive, there has been interest in the Symmetrical Measurement Equation (SME) approach to MTT. An SME tracker subsumes the data association step by defining a measurement equation where the ordering of measurements is no longer required. However, the acceptance of the SME approach has been curtailed by a lack of stability in the Extended Kalman Filter (EKF) and the ability of the SME track to properly maintain tracks during track crossings. Tweaking the EKF parameter can improve stability and track maintenance, but we show that an extension to the EKF filter, the Proportional-integral EKF, produces superior results. Integral action improved track maintenance at crossings, reducing the incidence of track switching by more than an order of magnitude. The SME tracker now shows promise of providing superior stability and reduced estimation error while tracking multiple, crossing, and maneuvering targets.

While single model filters are sufficient for tracking targets having fixed kinematic behavior, maneuvering targets require the use of multiple models. Jump Markov linear systems whose parameters evolve with time according to a finite state-space Markov chain, have been used in these situations with great success. However, it is well-known that performing optimal estimation for JMLS involves a prohibitive computational cost exponential in the number of observations. Many approximate methods have been proposed in the literature to circumvent this including the well-known GPB and IMM algorithms. These methods are computationally cheap but at the cost of being suboptimal. Efficient off- line methods have recently been proposed based on Markov chain Monte Carlo algorithms that out-perform recent methods based on the Expectation-Maximization algorithms. However, realistic tracking systems need on-line techniques. In this paper, we propose an original on-line Monte Carlo filtering algorithm to perform optimal state estimation of JMLS. The approach taken is loosely based on the bootstrap filter which, wile begin a powerful general algorithm in its original form, does not make the most of the structure of JMLS. The proposed algorithm exploits this structure and leads to a significant performance improvement.

For preliminary prediction of track accuracy in exoatmospheric applications, a common approach is to apply simple constant-velocity solutions separately in the range direction and two perpendicular crossrange directions. In typical radar applications, however, the range measurement errors are much smaller than the crossrange errors, and this approach neglects the important coupling effect by which the crossrange tracking accuracy, in the plane of motion of the line of sight, may be greatly improved due to the highly accurate range measurements. This paper demonstrates the extent of this effect by developing analytic solutions for the applicable Cramer-Rao bounds in the continuous- measurement case, which are then verified by numerical propagation of covariance matrices for the more realistic discrete-measurement case. A set of simple approximate expressions are presented, which are useful for prediction of tracking performance.

The problem of tracking point targets moving in a group, or features of an extended object, is formulated via a general two component model. An example involving translation, scaling, rotation and pattern distortion is presented. It is assumed that measurements of the points are unlabelled, which, together with a significant clutter level, leads to measurement association uncertainty. A Bayesian bootstrap filter is used to implement a nonlinear, multiple hypothesis, recursive estimator.

Active Defense with non-nuclear interceptors against Theater or Tactical Ballistic Missiles TBM will become feasible during the next few years. The efficient operation of lower or upper layer weapon systems mainly depend on a relatively accurate cue from satellites and/or long range radars. Geostationary satellites are able to observe the boost phase of TBM and the IR sensor can be designed to provide a rather accurate boost end state vector from which a sufficiently precise prediction of the TBMs free flight trajectory can be derived. Different search strategies for upper and lower layer weapon system radars have to be applied in order to achieve a high detection range gain compared to the autonomous mode of search. The paper will present a simulation approach which is used to evaluate the advantage of dedicated TBM Early Warning IR sensors on GEO satellites.

In a previous paper, we described AEW advanced tracking algorithm development and showed performance results for straight line targets. In this paper we show modifications of the tracking and association algorithms necessary to track the remaining 100 maneuvering targets of the 120 targets scenario. As part of ongoing AEW advanced tracking algorithm development activities, a novel approach has been developed that utilizes the Unbiased Coordinate Conversion Filter developed by Bar-Shalom et al for range and azimuth angle processing from the radar with a standard EKF for rdot and angular measurements from other sensors identified as a combined Kalman filter. We also incorporated fading memory into the filter to make it responsive to target maneuvers.

The federated filter is a near globally optimal distributed estimation method based on rigorous information-sharing principles. It is applied here to multi-perform target tracking systems where platform-level target tracks are fused across platforms into global tracks. Global track accuracy is enhanced by the geometric diversity of measurements from different platforms, in addition to the greater number of measurements. On each platform, the federated filter employs dual platform-level filters (PFs) for each track. The primary PFs are locally optimal, and contain all the information gathered from the platform track sensors. The secondary PFs are identical except that they contain only the incremental track information gained since the last fusion cycle. On each platform, global track solutions are near globally optimal because they receive only new tracklet information from the onboard and off-board PFs, and don to re-use old platform-level information. Logistically, platforms can operate autonomously with no need for synchronized operations or master/slave designations; the architecture is completely symmetric. Platforms can enter or leave the group with no changes in other global trackers. Communications bandwidth is minimal because global tracks need not be shared. The paper describes the theoretical basis of the federated fusing filter, the related data association functions, and preliminary simulation results.

This paper examines the determination of sensor biases by comparing multiple track outputs from spatially disparate platforms. In many cases, this can be achieved using least- square methods, minimizing the summed-squares of track differences. This method, while simple, is shown to provide valuable information converging the observability of a given type of bias. Observability is here defined as permitting the determination of a unique value for that bias, for each sensor in a pair, given sufficient measurements pairs. For the purposes of this study, the types of bias are divided into three main classes: measurement, own-position and alignment. In general, only members of the first group are individually observable, while for the other two classes observable, while for the other two classes observable combinations of the individual biases can be derived. The least-squares method also provides an effective method for assessing observability in practice, and in determining which estimated biases are most reliable in such circumstances.

Adding new sensor metric information into a data fusion process does not always improve performance and can sometimes produce poorer results. References 1 and 2 used examples to show that - in some instances and contrary to expectation - adding new information resulted in poorer rather than improved performance, even though the information itself was correct. Correct is being used here to describe data that may be in error because of sensor deficiencies but whose error characteristics are accurately described and known to the fusion process. In other words, the fusion process is not being lied to be misrepresentation of the data quality. In this sense, an individual data point may be inaccurate, but the fusion process is capable of properly weighting that point in an optimal sensors that its statistical inaccuracy does not damage the final product any more than a data point from a better sensor that has less statistical inaccuracy. In a multiple-sensor fusion process, these kinds of result have been cited as reasons for not using data from poorer quality sensors for fear of diluting the performance of the better quality sensors. This paper explores the counterintuitive findings for these referenced examples and evaluates under what conditions lesser quality sensor or sensor that mistakenly overestimate their own data quality should be allowed to contribute to a sensor fusion process.

Because there may be misassociations, performance evaluation of target tracking is complex due to ambiguities that can cause confusion about which target a track is following. This paper presents a methodology to deal with these ambiguities so that performance metrics can be used to regulate performance. The emphasis of this paper is on the first step of a two-step methodology, namely, the decision on which target to compare to a track so that measures of performance can be computed. However, no single methodology may be best for all the various types of tracking systems. In addition, no one methodology may be best for all the performance metrics that are appropriate to a specific system. Therefore, four methodologies are described to permit selection of the appropriate methodologies for a specific tracking system. Also discussed is the difference between the error characteristics of signal detection and target tracking. Target tracking exhibits not only the classical Type I and Type II errors but also another error type that is identified and referred to as Type III errors.

The integration of multiple sensors for target tracking is complex but has the potential to provide very accurate state estimates. For most applications, each sensor provides its information to a central location where the integration is performed and the resulting composite track can be very accurate when compared to the individual sensor tracks. This composite track has the potential to provide enhanced system decisions and targeting information not otherwise available. However, sensor bias can severely degrade composite tracking performance when it is not properly considered. This paper presents algorithms and simulation result for the composite tracking of maneuvering targets through the use of multisensor-multisite integration in the presence of sensor residual bias.

The well-known probabilistic data association (PDA) filter handles the uncertainty in measurement origins inherent in tracking-in-clutter problems by using a probabilistically weighted sum of all measurements in the gate. In fact, the measurements in the gate may or may not include the one originated from the target. As such, two hypothetical models can be set up, corresponding to the events that the target measurements is and is not in the gate, respectively. This paper present an approach that integrates the PDA filter with the multiple-order method in a coherent manner based on the use of the above two hypothetical models. It is shown theoretically that the standard PDA filter is a special case of the first-order Generalized Pseudo Bayesian algorithm in the proposed formulation using a particular set of model transition probabilities. It is then proposed to adopt the superior interacting multiple-model architecture in this new formulation to improve the performance. The new algorithm is capable of achieving better performance by tuning the transition probabilities at a computational complexity comparable to that of the PDA filter. Simulation results are provided.

The measurements of the two closely-spaced targets will be merged when the target echoes are not resolved in angle, range, or radial velocity. The modified Cramer Rao lower bound (CRLB) is given for monopulse direction-of-arrival (DOA) estimation for two unresolved Rayleigh targets and used to give insight into the antenna boresight pointing. A monopulse processing technique is given for DOA estimation of two unresolved Rayleigh targets. The Nearest Neighbor Joint Probabilistic Data Association Algorithm is extended to include the possibility of merged monopulse measurements of Rayleigh targets. The monopulse signals are incorporated into the data association as a feature to discriminate between merged and resolved measurements.

Seemingly benign transformations of data from one coordinate system to another can introduce bias errors resulting from nonlinearities in the underlying conversion equations. These biases, unless corrected, can affect the statistical fidelity of parameter estimates. This paper examines the effect of such biases on the 3D tracking of targets - specifically, the transformation from sensor measurements coordinates to Cartesian x-y-z coordinates. The standard approach to correcting for bias errors involves the adjustment of transformed measurements by the estimated biases. This estimation procedure, however, varies with the assumed relationship between the distributions governing the sensor measurement and the 'true' target position. At one extreme, the measurement can be considered fixed and the 'true' position varied; at the other extreme, 'truth' is fixed and the measurement varied. These two options are compared and contrasted with various hybrid alternatives that impose variable distributions on both the measurement and 'truth'.

Advances in image and signal processing permit implementation of sophisticated sensor fusion algorithms for tracking and target-identification. The polymorphic estimator employs a dual path architecture to accomplish these tasks simultaneously. It has been observed that there is an identification bias toward more agile targets. This bias can be overcome with higher quality sensors. In some cases more precise modeling of the target motion is an even better solution.

The problem of data association remains central in multitarget, multisensor, and multiplatform tracking. Lagrangian relaxation methods have been shown to yield near optimal answers in real-time. The necessity of improvement in the quality of these solutions warrants a continuing interest in these methods. A partial branch-and-bound technique along with adequate branching and ordering rules are developed. Lagrangian relaxation is used as a branching method and as a method to calculate the lower bound for subproblems. The result shows that the branch-and-bound framework greatly improves the solutions in less time than relaxation alone.

For multi-functional phased-array radar, basic sensor parameters are variable over a wide range and may be chosen individually for each track. To exploit this high flexibility, resource-saving techniques for combined target tracking and sensor control are demanded. This is particularly important for military air situations with agile targets.

This paper considers the problem of tracking dim unresolved ground targets and helicopters in heavy clutter with a ground based sensor. To detect dim targets the threshold must be set low which result in a large number of false alarms. The tracker typically uses the target dynamics to prevent the false tracks. The interesting aspect of this problem is that the targets may be or may become stationary. The tracks of stationary targets are difficult to discriminate from tracks formed by persistent false alarms.

This paper presents an interacting multiple model (IMM) filtering scheme for use in passive ranging with angle-only measurement data. The application is to an air-to-air multiple target tracking environment. The IMM implementation uses the Modified Spherical Coordinates (MSC) method for one of the filter models and a general acceleration model for the other filter. This method, in effect, provides maneuver detections to that range and range rate estimation errors from the MSC filter are minimized and an accurate covariance matrix is maintained to reflect these errors.

In this paper we consider the problem of tracking a maneuvering target in clutter. We apply an original on-line Monte Carlo filtering algorithm to perform optimal state estimation. Improved performance of the resulting algorithm over standard IMM/PDAF based filters is demonstrated.

A method of estimating relative range rate for airborne targets based on projective geometry is presented, where robustness to noise and clutter is achieved using a wavelet representation. The first stage is to extract simple geometric bar features at multiple resolutions from local extrema in a wavelet transform. These are then refined in position, orientation, scale and aspect ratio to fit regions of the target, for examples wings and fuselage. Affine transformations are then found which map these geometric components between image frames. Finally a relationship between affine transformations and changes in 3D viewing aspect is used to estimate inter-frame range ratio. This method has the robustness of those based on first and second-order moments while retaining sufficient information to unambiguously identify an affine transformation between frames. The method does not depend on region segmentation to define an outline of the target, and is therefore insensitive to noise. Results are presented for 3D simulations of aircraft flying against a background of sky with clouds, where accurate range data are available. Range ratios are accurate over a wide range of 3D target orientations, but errors are more pronounced when new parts of the target become visible.

This paper deals with the design, choice, and comparison of model sets in the multiple-model (MM) approach to adaptive estimation. Most representative problems of model-set choice and design are considered. As the basis of model-set choice and design, criteria for model-set comparison and choice based on base-state estimation, mode estimation, mode identification, hybrid-state estimation, and hypothesis testing are presented first. Several computationally efficient and easily implementable solutions of the model- set choice problems based on sequential hypothesis tests are presented. Some of these solutions are optimal. Their effectiveness is verified via simulation. How these criteria and result can be used for model-set design is demonstrated via several examples. It is also demonstrated how a probabilistic model of possible scenarios can be constructed.

This paper deals with the problem of detection and tracking of low observable small-targets from a sequence of IR images against structural background and non-stationary clutter. There are many algorithms reported in the open literature for detection and tracking of targets of significant size in the image plane with good results. However, the difficulties of detecting small-targets arise from the fact that they are not easily discernable from clutter. The focus of research in this area is to reduce the false alarm rate to an acceptable level. Triple Temporal Filter reported by Jerry Silverman et. al., is one of the promising algorithms in this are. In this paper, we investigate the usefulness of Max-Mean and Max-Median filters in preserving the edges of clouds and structural backgrounds, which helps in detecting small-targets. Subsequently, anti-mean and anti-median operations result in good performance of detecting targets against moving clutter. The raw image is first filtered by max-mean/max-median filter. Then the filtered output is subtracted from the original image to enhance the potential targets. A thresholding step is incorporated in order to limit the number of potential target pixels. The threshold is obtained by using the statistics of the image. Finally, the thresholded images are accumulated so that the moving target forms a continuous trajectory and can be detected by using the post-processing algorithm. It is assumed that most of the targets occupy a couple of pixels. Head-on moving and maneuvering targets are not considered. These filters have ben tested successfully with the available database and the result are presented.

Range deception, such as range-gate-pull-off (RGPO) is a common electronic countermeasure (ECM) technique used to defeat or degrade tracking radars. Although a variety of heuristic approaches/tricks have been proposed to mitigate the impact of this type of ECM on the target tracking algorithms, none of them involve a systematic means to reject the countermeasure signals. This paper presents a general and systematic approach, called Decomposition and Fusion (DF) approach, for target tracking in the presence of range deception ECM and clutter. It is effective against RGPO, range-gate-pull-in, and range false target ECM techniques for a radar system where the deception measurements have virtually the same angles as the target measurement. This DF approach has four fundamental components: (a) decomposing the validated measurements by determination of range deception measurements using hypothesis testing; (b) running one or more tracking filters using the detected range deception measurements only; (c) running a conventional tracking-in-clutter filter using the remaining measurements; (d) fusing the tracking filters by a probabilistically weighted sum of their estimates. Several algorithms within the DF approach are discussed.

We present a new method for detecting subpixel targets from hyperspectral images using wavelet transform. In this paper, we focus on cases where the spectral information is available and the observed target signatures are close to that of the background. In the present approach, instead of using the spectral information directly, we wavelet transform the spectrum of each spatial pixel and perform the analysis in the wavelet domain. The signatures of the true target and the background contaminants can be well separated in the wavelet domain if the spectral signature of the target is quasi-localized in the spectral domain. This paper exploits this concept to detect weak, subpixel targets.

Since the SCUD launches in the Gulf War, theater ballistic missile (TBM) systems have become a growing concern for the US military. Detection, fast track initiation, backfitting for launch point determination, and tracking and engagement during boost phase or shortly after booster cutoff are goals that grow in importance with the proliferation of weapons of mass destruction. This paper focuses on track initiation and backfitting techniques, as well as extending some earlier results on tracking a TBM during boost phase cutoff. Results indicate that Kalman techniques are superior to third order polynomial extrapolations in estimating the launch point, and that some knowledge of missile parameters, especially thrust, is extremely helpful in track initiation.

A reality faced in the practical application of signal detection is the inexact statistical knowledge of the underlying random processes. Accordingly, it is often desirable for a detector to possess robustness. In this paper, we review how the concept of manifold slope can be employed to admit the measurement of robustness thus allowing the degree of robustness to be a factor in the design of the signal detector. We then present new results that show how certain nonstandard decision regions can result in what we term 'negative boundaries' which have the potential to enhance robustness. An example of this approach is provided and the results compared to the classical Huber approach for robust detection.

The results of the experimental investigations concerning the spectral and polarization characteristics of Doppler radar water surface echo in X- and Ka radiowave bands are represented. Both the sea radar returns and shallow basin radar returns are being considered. The main attention is devoted to research the time-frequency distributions of such signals for the purpose of identification of different states of nonstationarity of radar sea echo for example the state of 'spikes' and 'pauses'. The high order spectral estimation is employed to obtain the trajectories of spectral components for research the dynamical characteristics of Doppler spectra. The two-dimensional auto-correlation function of such trajectories give opportunity to obtain the distinctive structure the properties of which can be made use for above mentioned identification. To research the dynamic properties of separate spectral lines in time-frequency distributions the signal processing algorithm, making possible to reveal the distinctive trajectories for estimation of their basic attributes is made use. The revealing of trajectory is fulfilled with use of the neural network algorithm.

The proposed recurrent and accurate and economical algorithms for digital filtering, which are based on 1D and 2D convolutions in the different difference formats are adduced in the paper. Delta modulation belong to these formats too. The methods for choosing of parameters for DM are developed. Due to the small word length, simplicity of operations and ease of their realizations, the mentioned convolutions are expedient for realization in digital filters based on the multiprocessor multipipeline structures, in particular, on homogeneous computing environments and neural structures.

Of particular interest are efficient algorithms of detection and acquisition of small targets with low signal-to-clutter ratio (SCR) for reality applications. Some algorithms like matched filtering (MF), dynamic programming (DPA) and multistage hypothesis test have been proposed. Although some experiments proved that these algorithms have good target detection performance, it is difficult to implement by hardware because of more computational requirement. In this paper one simple; and useful algorithm of the detection and acquisition for small targets with low SCR is presented. Input image is preprocessed with using background normalization and background removal, and SCR of targets of difference value image outputted in improved. After image segmentation with an adaptive threshold is finished, the target detection is performed by continuous multi-frame detection method using spatial-temporal continuous motion features of targets. The targets are identified with multistage continuous filtering and a main target selected is acquired and tracked with the window. Experiments proved that the algorithm could detect small targets with SCR equal to 3 in cloud background. The method has been used for IR tracking system in field and good tracking performance is acquired.

The ground-target tracking with dual-image sensor is discussed in this paper. The mission scenario is that fighter and helicopter track tactical ground targets such as tank, armored car and articulated vehicle. The airborne system which using a video camera and an 8-12 micrometers IR imager and processing with parallel DSP can track and identify ground-target in real time.

It is a challenging problem to track point source targets and small extending ones in clutter. Only using single-image sensor is usually not adequate to meet the requirement of a multi-target tracking (MTT) system. Carrying out the measurement through two IR waveband image sensors, and fusing efficiently measurement data of both sensors, we can track small extending targets precisely and improve the performance of the MTT system. In this paper, the multi- target tracking method based on 3-5 micrometers and 8-12 micrometers IR image sensors is introduced. The mission scenario is thought as tracking air target. The algorithm of the target image centroid estimator is discussed. In order to improve the precision of target tracking, the data fusion algorithm with Kalman filter is described for the target vector feature. THe simulation results indicate that the target tracking precision can be improved by tracking small extending targets with dual-IR image sensors.

The concept of using the best hypothesis in the minimum mean square error (MMSE) sense is introduced in this paper to provide alternative data association algorithms for tracking single or multiple targets with data from one or more senors. The concept of using the best hypothesis in estimation is also applied to the task of multiple model filtering. The motivation for using the estimate based on the single best hypothesis in the MMSE sense are tow fold. First, there are situations where there is a natural preference to make hard decisions rather than soft decisions. Secondly given that a state estimate is based on a single hypothesis, there is the desire to minimize the mean square of the estimation invovling hypotheses due to discrete possibilities, the traditional MMSE criterion leads to so called soft decisions that may not be appropriate for an interceptor with a small region of lethality while, in contrast, hard decision might increase the probability of kill. Also in processing feature for use in target typing, soft decisions may degrade performance more than would a reasonable hard decision algorithm. While the best hypothesis method may be preferred for certain applications, the improved performance might be at the expense of increased processing load. Since the capability of available processors is increasing rapidly, emphasis can be expected to elan toward algorithms that take advantage of this enhanced capability to provide improved performance based on the specific needs of each application. The emphasis of this paper is on algorithms for data and track association and multiple model filtering using single frame association methods but these methods can be extended to multiple data frame approaches.