This PDF file contains the front matter associated with SPIE Proceedings Volume 7445, including the Title Page, Copyright information, Table of Contents, Introduction, and the Conference Committee listing.

Multiple-Input Multiple-Output (MIMO) radars are a new generation of radar systems that bring with them many benefits compared to the traditional phased-array radars. This paper discuses localization techniques for multiple targets when a MIMO radar is used as a measurement tool. A multiple hypotheses-based approach is used to estimate parameters of targets from raw measurements. Received amplitudes and associated range bins are taken as raw measurements. The multiple hypothesis-based method is implemented in two steps. First, hypotheses are initialized using the fist q pairs of transmitters and receivers. Then, a sequential method is applied to initial hypotheses to find final estimates of targets. A comparison is also made between multistatic and MIMO radars for target detection and localization via simulations. The effect of putting threshold on raw data is taken into consideration in both detecting and localizing targets for multistatic radars. Finally, simulation results confirm the superiority of MIMO radars for multiple target localization.

The Interacting Multiple Model (IMM) estimator has been proven to be effective in tracking agile targets. Smoothing or retrodiction, which uses measurements beyond the current estimation time, provides better estimates of target states. Various methods have been proposed for multiple model smoothing in the literature. In this paper, a new smoothing method, which involves forward filtering followed by backward smoothing while maintaining the fundamental spirit of the IMM, is proposed. The forward filtering is performed using the standard IMM recursion, while the backward smoothing is performed using a novel interacting smoothing recursion. This backward recursion mimics the IMM estimator in the backward direction, where each mode conditioned smoother uses standard Kalman smoothing recursion. Resulting algorithm provides improved but delayed estimates of target states. Simulation studies are performed to demonstrate the improved performance with a maneuvering target scenario. The comparison with existing methods confirms the improved smoothing accuracy. This improvement results from avoiding the augmented state vector used by other algorithms. In addition, the new technique to account for model switching in smoothing is a key in improving the performance.

Overhead persistent surveillance systems are becoming more capable at acquiring wide-field image sequences for long time-spans. The need to exploit this data is becoming ever greater. The ability to track a single vehicle of interest or to track all the observable vehicles, which may number in the thousands, over large, cluttered regions while they persist in the imagery either in real-time or quickly on-demand is very desirable. With this ability we can begin to answer a number of interesting questions such as, what are normal traffic patterns in a particular region or where did that truck come from? There are many challenges associated with processing this type of data, some of which we will address in the paper. Wide-field image sequences are very large with many thousands of pixels on a side and are characterized by lower resolutions (e.g. worse than 0.5 meters/pixel) and lower frame rates (e.g. a few Hz or less). The objects in the scenery can vary in size, density, and contrast with respect to the background. At the same time the background scenery provides a number of clutter sources both man-made and natural. We describe our current implementation of an ultrascale capable multiple-vehicle tracking algorithm for overhead persistent surveillance imagery as well as discuss the tracking and timing performance of the currently implemented algorithm which is aimed at utilizing grayscale electrooptical image sequences alone for the track segment generation.

This paper describes an empirical approach to characterizing and simulating sea clutter based on the identification and grouping of so called 'clutter events'. Clutter events are grouped based on azimuth width (or equivalently existence time). The groups are characterized with regards to mean spike amplitude and relative occurrence rate of events. The character of the azimuth amplitude profile within a group is further characterized in terms of associated amplitude probability distribution function (apdf), amplitude profile and variance. The multiparameter characterization is shown to be sufficiently robust to allow the simulation of a scene that exhibits not only a qualitative similarity to the real clutter but a demonstrable quantitative correspondence. When the cumulative distribution function (cdf) of the clutter simulated per the new approach is compared with that of the real sea clutter returns an excellent match is achieved. Thus the new simulation method is shown to be consistent with the simpler but widely used apdf characterization.

Passive Coherent Location (PCL) systems use existing commercial signals (e.g., FM broadcast, digital TV) as the illuminators of opportunity in air defence systems. PCL Sytems have many advantages such as low cost, covert operation and low vulnerability to electronic counter measures, over conventional radar systems. The main disadvantage of PCL systems is that the transmitter locations and the transmitted signals cannot be controlled. Thus, it is possible to have multiple transmitters that transmit the same signal/frequency inside the coverage region of the receiver. Thus, multiple measurements that originated from different transmitters and reflected by the same target will be received. Even though using multiple transmitters will facilitate better estimates of the target states due to spatial diversity, one cannot use these measurements without resolving transmitter and measurement origin uncertainties. This adds another level of complexity to the standard data association problem where the uncertainty is only in measurement origins. That is, there are two uncertainties that need to be resolved in order to track multiple targets. One is the measurement-to-target association and the other is the measurement-to-transmitter association. In this work, a tracking algorithm is proposed to track multiple targets using PCL systems with the above data association uncertainties. The efficiency of the proposed algorithm is demonstrated on realistically simulated data.

We propose a novel algorithm for automatic aircraft classification. The proposed method makes numerical equivalents to shape, size and other aircraft features as critical criteria to constitute the algorithm for their correct classification. This method uses Inverse Synthetic Aperture Radar (ISAR) aircraft images that are making maneuvers that introduce aircraft rotation relative to the radar station. By means of analyzing the shape of the radar pulse and Doppler shifts that are caused by rotation of the aircraft, an image of the aircraft shape can be constructed. We computer simulated five different categories of ISAR images. We tested the proposed classification algorithm on these five categories and on two more categories taken from the Internet. One aircraft model is simulated and the other one is a real sequence with much added noise. All seven different aircraft models are flying a holding pattern. We investigated where in the holding patterns ISAR reflections made it possible to identify each category of aircraft. Our experimental results demonstrate that in most parts of the holding pattern the category of the aircraft can be successfully identified. The performed tests show that the proposed algorithm appears to be noise resistant.

Video-based tracking of small targets in a dense environment of clutter is very difficult, because the image resolution of the target is too low to provide reliable information for matching, and in turn the clutter generates a large number of false positive matches and distractions. Most traditional methods attempt to oppose the target to the environment, and are thus confronted in handling the enormous distractions. In fact, a target is rarely isolated and independent to the environment, e.g., when persistent disturbances are present in the vicinity of the target. Therefore, there may exist some objects that exhibit short-term or even longer-term motion correlation to the target. They constitute a very useful spatial contexts of the target. Thus, taking the advantage of the contextual information in an efficient way can improve the robustness of target tracking, as the spatial contexts provide extra constraints in target matching and additional verification in data association. This paper presents a new approach of context-aware tracking for small targets, in which a set of motion-correlated auxiliary objects are automatically discovered on-the-fly. The image region of one such auxiliary object generates a specific spatial context of the target, and leads to an individual contextual constraint to the motion of the target. Under the small motion assumption on two consecutive frames, these individual contextual constraints have linear forms. The collection of all such individual contextual constraints gives a contextual system, based on which the target motion can be accurately estimated so that the association of the target over consecutive image frames can be reliably constructed. This new approach is computationally efficient. Extensive experiments on real test video sequences show the effectiveness and efficiency of the proposed approach.

Detection of unusual trajectories of moving objects can help in identifying suspicious activity on convoy routes and thus reduce casualties caused by improvised explosive devices. In this paper, using video imagery we compare efficiency of various techniques for incremental outlier detection on detecting unusual trajectories on simulated and real-life data obtained from SENSIAC database. Incremental outlier detection algorithms that we consider in this paper include incremental Support Vector Classifier (incSVC), incremental Local Outlier Factor (incLOF) algorithm and incremental Connectivity Outlier Factor (incCOF) algorithm. Our experiments performed on ground truth trajectory data indicate that incremental LOF algorithm can provide better detection of unusual trajectories in comparison to other examined techniques.

A method for ground-penetrating radar amplitude recovery of targets is presented in this paper. We use the migration processing for target imaging and the clean method to extract featured target points. Based on the featured target points, target reflection hyperbolas are calculated. Curve fitting is applied to get amplitude attenuation parameters. The amplitude is recovered after compensating the parameters to the raw data. We performed a field experiment with 5 sphere targets under water. Comparing to the conventional methods, our proposed method is more effective for target classification.

In this paper, we study a nonlinear bearing-only target tracking problem using four different estimation strategies and compare their performances. This study is based on a classical ground surveillance problem, where a moving airborne platform with a sensor is used to track a moving target. The tracking scenario is set in two dimensions, with the measurement providing angle observations. Four nonlinear estimation strategies are used to track the target: the popular extended and unscented Kalman filters (EKF/UKF), the particle filter (PF), and the relatively new smooth variable structure filter (SVSF). The SVSF is a predictor-corrector method used for state and parameter estimation. It is a sliding mode estimator, where gain switching is used to ensure that the estimates converge to true state values. An internal model of the system, either linear or nonlinear, is used to predict an a priori state estimate. A corrective term is then applied to calculate the a posteriori state estimate, and the estimation process is repeated iteratively. The performances of these methods applied on a bearing-only target tracking problem are compared in terms of estimation accuracy and filter robustness.

The paper presents an automatic approach for small moving object detection, categorization and unusual motion pattern signaling on camera feeds on sky background. The method uses a local blind deconvolution based foreground detector for small object mask and contour edge extraction, spatio-temporal localized histogram evaluation for object classification, and a hidden Markov model based evaluation for learning usual motions and signaling unusual motion patterns. The method is able to mask moving objects, fit them into learned categories and signal unexpected motion behavior.

The probability hypothesis density (PHD) filter is an estimator that approximates, on a given scenario, the multitarget distribution through its first-order multitarget moment. This paper presents two particles labeling algorithms for the PHD particle filter, through which the information on individual targets identity (otherwise hidden within the first-order multitarget moment) is revealed and propagated over time. By maintaining all particles labeled at any time, the individual target distribution estimates are obtained under the form of labeled particle clouds, within the estimated PHD. The partitioning of the PHD into distinct clouds, through labeling, provides over time information on confirmed tracks identity, tracks undergoing initiation or deletion at a given time frame, and clutter regions, otherwise not available in a regular PHD (or track-labeled PHD). Both algorithms imply particles tagging since their inception, in the measurements sampling step, and their re-tagging once they are merged into particle clouds of already confirmed tracks, or are merged for the purpose of initializing new tracks. Particles of a confirmed track cloud preserve their labels over time frames. Two data associations are involved in labels management; one assignment merges measurement clouds into particle clouds of already confirmed tracks, while the following 2D-assignment associates particle clouds corresponding to non-confirmed tracks over two frames, for track initiation. The algorithms are presented on a scenario containing two targets with close and crossing trajectories, with the particle labeled PHD filter tracking under measurement origin uncertainty due to observations variance and clutter.

The probability hypothesis density (PHD) and cardinalized PHD (CPHD) filters were introduced as approximations of the full multitarget Bayes detection and tracking filter. Both filters are based on the "standard" multitarget measurement model that underlies most multitarget tracking theory. That is, sensor measurements are presumed to be detections. Other sensors collect measurements that are not detections, and among the most important of these are superpositional sensors. A measurement collected by such a sensor is a sum of the real- or complex-valued signals generated by an unknown number of unknown targets present in the scene. This paper describes a theoretical extension of the CPHD filter concept to superpositional sensors.

The Probability Hypothesis Density (PHD) filter is a powerful new tool in the field of multitarget tracking. Unlike classical multi-target tracking approaches, such as Multiple Hypothesis Tracking (MHT), in each scan it provides a complete solution to multi-target state estimation without the necessity for explicit measurement-to-track data association. The PHD filter recursively propagates the first order moment of the multi-target posterior. This allows us to determine the expected number of targets as well as their state estimates at each scan. However, there is no implicit connection between the target state estimates in consecutive scans. In this paper, a new cluster-based approach is proposed for track labeling in the Sequential Monte Carlo (SMC i.e. particle filter based) PHD filter. The method associates a likelihood vector to each particle in the SMC estimate. This vector indicates the likelihood that the particle estimate belongs to each of the established target tracks. This likelihood vector is propagated along with the PHD moment and updated with the PHD function. By maintaining a set of associations from scan to scan, the new method provides a complete PHD solution for a multi-target tracking application over time. The method is tested on both clean and noisy multi-target tracking scenarios and the results are compared to some previously published methods.

The track repulsion effect induces track swapping in difficult target-crossing scenarios. This paper provides a simple analytical model for the probability of successful tracking in this setting. The model provides a means to quantify the degree-of-difficulty in target-crossing scenarios. We analyze model-based performance predictions for a range of scenario parameters. Additionally, we provide simulation results with a multi-hypothesis tracker that confirm the increased performance challenge in crossing target settings as the ambiguity persists longer, i.e. as the targets cross more slowly.

Probability Hypothesis Density (PHD) filter is a unified framework for multitarget tracking and provides estimates for a number of targets as well as individual target states. Sequential Monte Carlo (SMC) implementation of a PHD filter can be used for nonlinear non-Gaussian problems. However, the application of PHD based state estimators for a distributed sensor network, where each tracking node runs its own PHD based state estimator, is more challenging compared with single sensor tracking due to communication limitations. A distributed state estimator should use the available communication resources efficiently in order to avoid the degradation of filter performance. In this paper, a method that communicates encoded measurements between nodes efficiently while maintaining the filter accuracy is proposed. This coding is complicated in the presence of high clutter and instantaneous target births. This problem is mitigated using novel adaptive quantization and encoding techniques. The performance of the algorithm is quantified using a Posterior Cramer-Rao Lower Bound (PCRLB), which incorporates quantization errors. Simulation studies are performed to demonstrate the effectiveness of the proposed algorithm.

The "classical" PHD and CPHD filters presume the standard "small-target" detection model. This year, in a series of theoretical studies, I have derived new "second-generation" CPHD/PHD filters for various sensing conditions that cannot be described by the standard model. These are: (a) multisensor, (b) clutter estimation, (c) tracking in unknown clutter, (d) extended targets, (e) unresolved targets, and (f) superpositional sensors. A common factor underlies all of these derivations: the FISST multitarget calculus. It is possible, given that one already knows the correct "answer," to reverse engineer the classical PHD/CPHD filters and to extemporize some "elementary" means of deriving them. But only the multitarget calculus is guaranteed to result in theoretically rigorous formulas for new problems-i.e., those for which the answer is not known beforehand. I also announce an important new result: the multitarget state estimators used with the CPHD/PHD filters are Bayes-optimal.

In this paper, a new state estimation algorithm for estimating the states of targets that are separable into linear and nonlinear subsets with non-Gaussian observation noise distributed according to a mixture of Gaussian functions is proposed. The approach involves modeling the collection of targets and measurements as random finite sets and applying a new Rao-Blackwellised Approximate Conditional Mean Probability Hypothesis Density (RB-ACM-PHD) recursion to propagate the posterior density. The RB-ACM-PHD filter jointly estimates the time-varying number of targets and the observation sets in the presence of data association uncertainty, detection uncertainty, noise and false alarms. The proposed algorithm approximates a mixture Gaussian distribution with a moment-matched Gaussian in the weight update phase of the filtering recursion. A two dimensional maneuvering target tracking example is used to evaluate the merits of the proposed algorithm. The RB-ACM-PHD filter results in a significant reduction in computation time while maintaining filter accuracies similar to the standard sequential Monte Carlo PHD implementation.

This paper presents a real time system for tracking multiple ground moving targets in aerial video. The state of a target is described by its kinematics as well as shape and appearance features: the kinematics include location and velocity in an earth fixed coordinate system; the shape is described by the parameters of an ellipse; the appearance features consist of color histogram, color correlogram, edge matching and/or orientation correlation information. The target kinematics is represented by a constant velocity model and the shape and appearance features are represented by static models between two observation instances. The motion layers of elliptical shapes containing moving targets in stabilized video sequence are identified. The location and velocity in geospace and the corresponding covariances are computed for each target within a motion layer using the platform metadata. A k-best joint probabilistic data association (JPDA) algorithm updates the target kinematics, while an α-β filter updates the shape and appearance features. Additionally, the JPDA assignment cost matrix is formulated using the kinematics, the appearance features, and the target heading information. The k-best Hungarian algorithm is used to obtain the best assignments. The issues of target life cycle management and target splitting and merging are also addressed in our framework. The system has been tested and evaluated for vehicle tracking in sparse, medium, and dense traffic using aerial EO and IR videos.

This paper presents a novel method for tracking ground moving targets with a GMTI radar. To avoid detection by the GMTI radar, targets can deliberately stop for some time before moving again. The GMTI radar does not detect a target when the radial velocity (along the line-of-sight from the sensor) falls below a certain minimum detectable velocity (MDV). We develop a new approach by using state-dependent mode transition probabilities to track move-stop-move targets. Since in a real scenario, the maximum deceleration is always limited, a target can not switch to the stopped-target model from a high speed. Therefore, with the use of the stopped-target model, the Markov chain of the mode switching has jump probabilities that depend on the target's kinematic state. A mode transition matrix with zero jump probabilities to the stopped-target mode is used when the speed is above a certain "stopping" limit (above which the target cannot stop in one sampling interval, designated as "fast stage") and another transition matrix with non-zero jump probabilities to the stopped-target mode is used when the speed is below this limit (designated as "slow stage"). The stage probabilities are calculated using the kinematic state statistics from the IMM estimator and then used to combine the state-dependent mode transition probabilities (SDP) in the two different transition matrices. The experimental results show that the proposed algorithm outperforms previous methods.

We present a target tracking system for a specific sort of passive radar, that using a Digital Audio/Video Broadcast (DAB/DVB) network for illuminators of opportunity. The system can measure bi-static range and range-rate. Angular information is assumed here unavailable. The DAB/DVB network operates in a single frequency mode; this means the same data stream is broadcast from multiple senders in the same frequency band. This supplies multiple measurements of each target using just one receiver, but introduces an additional ambiguity, as the signals from each sender are indistinguishable. This leads to a significant data association problem: as well as the usual target/measurement uncertainty there is additional "list" of illuminators that must be contended with. Our intention is to provide tracks directly in the geographic space, as opposed to a two-step procedure of formation of tracks in (bi-static) range and range-rate space to fuse these onto a map. We offer two solutions: one employing joint probabilistic data association (JPDA) based on an Extended Kalman Filter (EKF), and the other a particle filter. For the former, we explain a "super-target" approach to bring what might otherwise be a three-dimensional assignment list down to the two dimensions the JPDAF needs. The latter approach would seem prohibitive in computation even with these; as such, we discuss the use of a PMHT-like measurement model that greatly reduces the numerical load.

The primary components of a target track are the estimated state vector and its error variance-covariance matrix (or simply the covariance). The estimated state indicates the location and motion of the target. The track covariance is intended to indicate the uncertainty or inaccuracy of the target state estimate. The covariance is computed by the track processor and may or may not realistically indicate the inaccuracy of the state estimate. Covariance Consistency is the property that a computed variance-covariance matrix realistically represents the covariance of the actual errors of the estimate. The computed covariance of the state estimation error is used in the computations of the data association processing function and the estimation filter; consequently, degraded track consistency might cause misassociations (correlation errors) and degraded filter processing that can degrade track performance. The computed covariance of the state estimation error is also used by downstream functions, such as the network-level resource management functions, to indicate the accuracy of the target state estimate. Hence, degraded track consistency can mislead those functions and the war fighter about the accuracy of each target track. In the development of target trackers, far more attention has been given to improving the accuracy of the estimated target state than in improving the track covariance consistency. This paper addresses covariance compensation to reduce the degradation of consistency due to potential misassociations in measurement fusion using single-frame data association. The compensation approach used is also applicable to other fusion approaches and to tracking with data from a single sensor. This paper also shows how this compensation approach can be applied to a wide variety of data association algorithms.

The coordinated use of multiple distributed sensors by network communication has the potential to substantially improve track state estimates even in the presence of enemy countermeasures. In the modern electronic warfare environment, a network-centric tracking system must function in a variety of jamming scenarios. In some scenarios hostile electronic countermeasures (ECM) will endeavor to deny range and range rate information, leaving friendly sensors to depend on passive angle information for tracking. In these cases the detrimental effects of ECM can be at least partially ameliorated through the use of multiple networked sensors, due to the inability of the ECM to deny angle measurements and the geometric diversity provided by having sensors in distributed locations. Herein we demonstrate algorithms for initiating and maintaining tracks in such hostile operating environments with a focus on maximum likelihood estimators and provide Cramer-Rao bounds on the performance one can expect to achieve.

This paper considers the problem of dynamic residual bias estimation in the presence of measurement association uncertainty using common targets of opportunity under a decentralized information processing architecture i.e. independent trackers at each sensor. This is done by extending the scope of the synchronous version of the bias estimation algorithm presented by Lin, Bar-Shalom and Kirubarajan in "Multisensor-Multitarget Bias Estimation for General Asynchronous Sensors" to develop approaches to bias estimation in the presence of measurement association uncertainty. We consider the computational complexity and the sensor-to-fusion-center communication requirements of each of these approaches and compare their simulated performance in terms of RMSE and consistency. Though the simulations are performed with synchronous polar measurements having additive biases, the algorithm may easily be extended to the case with asynchronous measurements in other coordinate systems having both additive and multiplicative biases.

In this paper, several nonlinear filters (EKF/CMKF/CMEKF, UKF and PFs) are compared using real datasets and simulations based on two representative radar datasets. The first dataset was collected from an air traffic control (ATC) radar experiment with several aircraft. The second dataset was recorded from a high frequency surface wave radar (HFSWR) trial that was characterzed by a very long integration time and a limited set of manoeuvre types. RMSE, NEES and NIS are used as measures of performance. Comments on the performance, computational requirements of the nonlinear filters, practical modelling and filter tuning issues for the two types of radars are also presented.

We solve the fundamental and well known problem in particle filters, namely "particle collapse" or "particle degeneracy" as a result of Bayes' rule. We do not resample, and we do not use any proposal density; this is a radical departure from other particle filters. The new filter implements Bayes' rule using particle flow rather than with a pointwise multiplication of two functions. We show numerical results for a new filter that is vastly superior to the classic particle filter and the extended Kalman filter. In particular, the computational complexity of the new filter is many orders of magnitude less than the classic particle filter with optimal estimation accuracy for problems with dimension greater than 4. Moreover, our new filter is two orders of magnitude more accurate than the extended Kalman filter for quadratic and cubic measurement nonlinearities. We also show excellent accuracy for problems with multimodal densities.

This paper discusses the application of multiple hypothesis tracking (MHT) to the processing of ground target data collected with a long range surveillance radar. A key element in the successful tracking of ground targets is the use of road networks. Thus, the paper begins with an overview of the alternative approaches that have been considered for incorporating road data into a ground target tracker and then it gives a detailed description of the methods that have been chosen. The major design issues to be addressed include the manner in which road filter models are included into a Variable-Structure Interacting Multiple Model (IMM) filtering scheme, how the road filter models are chosen to handle winding roads and intersections, and the tracking of targets that go on and off-road. Performance will be illustrated using simulated data and real data collected from a large surveillance area with a GMTI radar. The area considered contains regions of heavy to moderate target densities and clutter. Since the real data included only targets of opportunity (TOO), it was necessary to define metrics to evaluate relative performance as alternative tracking methods/parameters are considered. These metrics are discussed and comparative results are presented.

Collision warning systems are used for commercial air traffic to provide pilots with an extra layer of situational awareness (as well as avoidance actions). The recent surge in the use of unmanned aerial vehicles (UAV) means the design of a collision warning system for UAVs is increasing in importance. This paper deals with the design for consistency of a Passive Collision Warning System (PCWS) which utilizes low resolution infrared (IR) cameras mounted to the airframe. The lack of range information, as well as the unknown measurement noise statistics, make tracking and the decision for collision warning difficult. Of the utmost importance is the estimation of the measurement noise variance of the sensor and the consistency of the resulting tracking filter. The proposed PCWS adaptively estimates this noise variance. The resulting system was found to provide consistent tracking filters as well as accurate estimates of the angular velocity of detected targets, verified through a number of test flights with an aircraft passing both over a stationary camera as well as across its field of view. Subsequently, the filter was modified for use on a light aircraft in conjunction with an inertial navigation system.

Existing unbiased converted measurement Kalman filters (CMKF) may still give biased estimates under some situations. The covariance of the converted measurement conditioned on the measurements is a noisy stochastic process with strong correlation with the measurement noise; therefore, the filter gain of the CMKF also becomes dependent on the measurement noise. Consequently the measurement noise weighted by the noise-dependent filter gain will no longer be zero mean, hence it can cause the CMKF to become biased. By using the converted measurement covariance at the previous time instead of the one at the current time, the filter gain of the CMKF is decorrelated from the measurement noise, which makes the weighted innovations zero mean. Simulation results show that the proposed CMKF with decorrelated measurement covariance runs with no bias in all situations.

We have investigated more than 17 distinct methods to approximate the gradient of the loghomotopy for nonlinear filters. This is a challenging problem because the data are given as function values at random points in high dimensional space. This general problem is important in optimization, financial engineering, quantum chemistry, chemistry, physics and engineering. The best general method that we have developed so far uses a simple idea borrowed from geology combined with a fast approximate k-NN algorithm. Extensive numerical experiments for five classes of problems shows that we get excellent performance.

Laser pointing systems for small targets are mainly confronted with two pointing errors, jitter and boresight, arising due to atmospheric turbulence, mechanical vibrations and errors in optical alignment. Knowledge of these parameters provides information about the quality of the pointing and tracking system. In the past, several techniques have been investigated to estimate these parameters from returned laser signals. These include the key ratio technique, the chi-squared method and the maximum likelihood (ML) estimation technique. These techniques have been studied in the literature for Gaussian irradiance profiles. In particular, the ML estimation technique has been found to provide excellent results. In this paper, we extend the ML estimation technique from Gaussian profiles to near-Gaussian irradiance profiles. Our results show that the modified estimator performs much better than a direct application of the original ML estimator.

In this paper, methods of tracking multiple targets in non-homogeneous clutter background is studied. In many scenarios, after detection process, measurement points provided by the sensor (e.g., sonar, infrared sensor, radar) are not distributed uniformly in the surveillance region. On the other hand, in order to obtain accurate results, the target tracking filter requires information about clutter's spatial density. Thus, non-homogeneous clutter point spatial density has to be estimated based on the measurement point set and tracking filter's outputs. Also, due to the requirement of compatibility, it is desirable for this estimation method to be integrated into current tracking filters. In this paper, a recursive maximum likelihood method and an approximated Bayesian method are proposed to estimate the clutter point spatial density in non-homogeneous clutter background and both will in turn be integrated into Probability Hypothesis Density (PHD) filter. Here, non-homogeneous Poisson point processes, whose intensity function are assumed to be mixtures of Gaussian functions, are used to model clutter points. The mean and covariance of each Gaussian function is estimated and used in the update equation of the PHD filter. Simulation results show that the proposed methods are able to estimate the clutter point spatial density and improve the performance of PHD filter over non-homogeneous clutter background.

Target classification based on the transferable belief model (TBM) is believed to be more robust than the Bayesian method. However, existing TBM classifier may forget over time the estimated prior information of the class. This paper proposes a recursive TBM classifier, which could combine the current basic belief assignment (BBA) of the class with the historic class information. Besides, feature mapping from the feature space to the class space, instead of the conventional converse mapping, is utilized to improve the performance of the recursive classifier. Simulation results reveal that the proposed TBM classifier eliminated the deficiency of existing TBM method and has more robust performance than the Bayesian classifier.

In this paper an algorithm for multipath-assisted multitarget tracking using multiframe assignment is proposed for initiating and tracking multiple targets using one or more transmitters and receivers. This algorithm is capable of exploiting multipath target returns from distinct propagation modes that are resolvable by the receiver. When resolved multipath returns are not utilized within the tracker, i.e., discarded as clutter, potential information conveyed by the multipath detections of the same target is wasted. In this case, spurious tracks are formed using target-originated multipath measurements, but with an incorrect propagation mode assumption. Integrating multipath information into the tracker (and not discarding it) can help improve the accuracy of tracking and reduce the number of false tracks. The challenge in improving tracking results using multipath measurements is the fusion of direct and multipath measurements from the common target. The problem will be considered in an environment with false alarms and missed detections. We propose a multiframe assignment technique to incorporate multipath information. The simulation results are presented to show the effectiveness of the proposed algorithm with an example of tracking ground targets.

To accommodate the flow of commerce, cargo inspection systems require a high probability of detection and low false alarm rate while still maintaining a minimum scan speed. Since objects of interest (high atomic-number metals) will often be heavily shielded to avoid detection, any detection algorithm must be able to identify such objects despite the shielding. Since pixels of a shielded object have a greater opacity than the shielding, we use a clustering method to classify objects in the image by pixel intensity levels. We then look within each intensity level region for sub-clusters of pixels with greater opacity than the surrounding region. A region containing an object has an enclosed-contour region (a hole) inside of it. We apply a region filling technique to fill in the hole, which represents a shielded object of potential interest. One method for region filling is seed-growing, which puts a "seed" starting point in the hole area and uses a selected structural element to fill out that region. However, automatic seed point selection is a hard problem; it requires additional information to decide if a pixel is within an enclosed region. Here, we propose a simple, robust method for region filling that avoids the problem of seed point selection. In our approach, we calculate the gradient Gx and Gy at each pixel in a binary image, and fill in 1s between a pair of x_{1 Gx(x1,y)=-1} and x_{2 Gx(x2,y)=1}, and do the same thing in y-direction. The intersection of the two results will be filled region. We give a detailed discussion of our algorithm, discuss the strengths this method has over other methods, and show results of using our method.

In this paper, the problem of detection, classification and tracking of highly manoeuvring boats in sea clutter is considered. The considered problem is challenging due to numerous inherent issues: abrupt direction changes, high level of false alarms, lowered detectability, group movement and re-grouping, among many others. The results of applying a proposed measurement extraction and estimation technique to a set of real data from DRDC-Ottawa trials using Ground Moving Target Indicator (GMTI) radar are described. Real radar data containing a small manoeuvring boat in sea clutter is processed using Expectation Maximization (EM) Gaussian Mixture Model (GMM) based estimation. A trial was undertaken to collect data against highly maneuvering speedboats in the sea. All the data were collected in the GMTI single-channel high-resolution spotlight mode. True data were collected using GPS recording equipment. Real data processing results are presented.

Passive Coherent Location (PCL) is a low-cost system for tracking of air targets clandestinely using illuminators of opportunity such as FM broadcast and digital TV. Due to an increased interest in PCL systems, researchers have been working on different configurations of available sources of opportunity and receivers capable of extracting plots from reflected signals of opportunity. The configuration can be multiple-transmitter-single-receiver or singletransmitter- multiple-receiver. Unlike standard radar systems, which can be optimized for detection probability and/or false alarm rate using different transmitted signals and adaptive thresholding, PCL systems are prone to poor detection due to low signal-to-noise (SNR). This leads to high clutter with low probability of detection of target of interest. In this work, we implement a multisensor-multitarget tracking system that fuses measurements from different PCL systems to improve tracking results. The benefits of the fusion are demonstrated using real data from NATA SET Panel 108 on PCL as well as simulated data.

Dual energy imaging is a technique whereby an object is scanned with X-rays of two interleaved energies to extract information about the object's atomic composition (Z). This technique is based on the fact that the X-ray absorption coefficient decreases with X-ray energy for low-Z materials, but begins to increase for high-Z materials due to the onset of pair production. Methods using the ratio of the attenuations of high-energy to low-energy images as an indicator of Z value have been proposed by several people. However, the statistical errors associated with the systems make those indicators unreliable. Methods that calculate the probability of high Z encounter a problem of what is the threshold probability to call high Z to minimize both miss and false alarm. We have developed an "adaptive regional masking" method that avoids the predicament of a single threshold. Our method is adaptive because the threshold for determining high Z varies adaptively in different regions on the image. The "mask" refers to the location of objects in the thickness map that mask to possible high-Z regions. Adaptive thresholding improves detection, while masking reduces false alarms. Test results show an increased accuracy of high-Z detection using this approach. In this paper, we discuss the approach and show some sample test results illustrating the effectiveness of the method.