This paper considers the problem of estimating the 3D states of a salvo of thrusting/ballistic endo-atmospheric objects using 2D Cartesian measurements from the focal plane array (FPA) of a single fixed optical sensor. Since the initial separations in the FPA are smaller than the resolution of the sensor, this results in merged measurements in the FPA, compounding the usual false-alarm and missed-detection uncertainty. We present a two-step methodology. First, we assume a Wiener process acceleration (WPA) model for the motion of the images of the projectiles in the optical sensor’s FPA. We model the merged measurements with increased variance, and thence employ a multi-Bernoulli (MB) filter using the 2D measurements in the FPA. Second, using the set of associated measurements for each confirmed MB track, we formulate a parameter estimation problem, whose maximum likelihood estimate can be obtained via numerical search and can be used for impact point prediction. Simulation results illustrate the performance of the proposed method.

Both the Maximum Likelihood Probabilistic Data Association (ML-PDA) track extractor and the Maximum Likelihood Probabilistic Multi-Hypothesis (ML-PMHT) track extractor are extended in this work to handle the scenario of two unresolved moving objects in the field of gravity. The original ML-PDA and ML-PMHT log-likelihood ratios are modified to use the probability that the objects being tracked are unresolved i.e., their measurements are merged. The performances of the modified ML-PDA and ML-PMHT, which we denote MLPDA-M and ML-PMHT-M (M for merged), respectively, are compared with those of the ML-PDA and the ML-PMHT in a notional scenario in which two moving objects appear initially unresolved to two space-based passive sensors observing them and become resolved first by one and then by both. Simulation results for the original track extractors and the modified track extractors are presented. While in many tracking situations the performances of the ML-PDA and ML-PMHT are indistinguishable (and the ML-PMHT therefore selected for its other features), this case of challenged resolution appears to be one situation where the more arduous ML-PDA ought to be favored. There does seem to be some reason to favor the “M” versions of both, but the results there are less compelling.

A passive ranging problem with elevation angle, azimuth angle and signal intensity measurements is presented and solved with a Maximum Likelihood (ML) estimator. The measurements used in the estimation are all obtained from a single passive sensor at a fixed location. The intensity measurement, which obeys the inverse square law w.r.t. the squared distance between the sensor and target has an unknown emitted energy that needs to be taken into account in the estimation problem. The Fisher Information Matrix (FIM) is investigated and used for observability testing. The simulation results from the scenarios considered prove the efficiency of the ML estimator.

This paper presents an algorithm, based on the Interacting Multiple Model Estimator, that can be used to track the state of kinematic point targets, moving in two dimensions, that are capable of making sharp heading maneuvers over short periods of time, such as certain ground vehicles moving in an open field. The targets are capable of up to 60 °/s turn rates, while polar measurements are received at 1 Hz. We introduce the Non-Zero Mean, White Noise Turn-Rate IMM (IMM-WNTR) that consists of 3 modes based on a White Noise Turn Rate (WNTR) kinematic model that contains additive, white, Gaussian turn rate process noises. Two of the modes are considered maneuvering modes, and they have opposite (left/right), non-zero mean turn rate input noise. The need for non-zero mean turn rate process noise is explained, and Monte Carlo simulations compare this novel design to the traditional (single-mode) White Noise Acceleration Kalman Filter (WNA KF) and the two-mode White Noise Acceleration/Nearly-Coordinated Turn Rate IMM (IMM-CT). Results show that the IMM-WNTR filter achieves better accuracy and real-time consistency between expected error and actual error as compared to the (single-mode) WNA KF and the IMM-CT in all simulated scenarios, making it a very accurate state estimator for targets with sharp coordinated turn capability in 2D.

We are tracking multiple targets based on noisy measurements. The targets are labeled, the measurements are unlabeled, and the association of measurements to targets is unknown. Our goal is association-free tracking, so the associations will never be determined as this is costly and impractical in many scenarios. By employing a permutation-invariant and differentiable point set distance measure, we derive a modified association-free multi-target measurement equation. It maintains the target identities but is invariant to permutations in the unlabeled measurements. Based on this measurement equation, we derive an efficient sample-based association-free multi-target Kalman filter. The proposed new approach is straightforward to implement and scalable.

The ability to detect and unambiguously follow all moving entities in a state-space is important in multiple domains both in defence (e.g. air surveillance, maritime situational awareness, ground moving target indication) and the civil sphere (e.g. astronomy, biology, epidemiology, dispersion modelling). However, tracking and state estimation researchers and practitioners have difficulties recreating state-of-the-art algorithms in order to benchmark their own work. Furthermore, system developers need to assess which algorithms meet operational requirements objectively and exhaustively rather than intuitively or driven by personal favourites. We have therefore commenced the development of a collaborative initiative to create an open source framework for production, demonstration and evaluation of Tracking and State Estimation algorithms. The initiative will develop a (MIT-licensed) software platform for researchers and practitioners to test, verify and benchmark a variety of multi-sensor and multi-object state estimation algorithms. The initiative is supported by four defence laboratories, who will contribute to the development effort for the framework. The tracking and state estimation community will derive significant benefits from this work, including: access to repositories of verified and validated tracking and state estimation algorithms, a framework for the evaluation of multiple algorithms, standardisation of interfaces and access to challenging data sets. Keywords: Tracking,

Bias estimation for multiple passive sensors using common targets of opportunity has been researched extensively. However, the proposed solutions required the use of multiple (two or more) passive sensors. In order to remove this constraint, we provide in this paper a new methodology using a single exoatmospheric target of opportunity seen in a single satellite borne sensor’s field of view to estimate the sensor’s biases simultaneously with the state of the target. The satellite is equipped with an optical sensor that provides the Line Of Sight (LOS) measurements of azimuth and elevation to the target. The evaluation of the Cram´er-Rao Lower Bound (CRLB) on the covariance of the bias estimates, and the statistical tests on the results of simulations show that this method is statistically efficient.

This paper is the sixth in a series aimed at weakening the independence assumptions that are typically presumed in multitarget tracking. Earlier papers investigated Bayes …lters that propagate the correlations between two evolving multitarget systems. Last year at this conference we attempted to derive PHD …lter-type approximations that account for both spatial correlation and cardinality correlation (i.e., correlation between the target numbers of the two systems). Unfortunately, this approach required heuristic models of both clutter and target appearance in order to incorporate both spatial and cardinality correlation. This paper describes a fully rigorous approach- provided, however, that spatial correlation between the two populations is ignored and only their cardinality correlations are taken into account. We derive the time-update and measurement-update equations for a CPHD …lter describing the evolution of such correlated multitarget populations.

This paper is the seventh in a series aimed at weakening the independence assumptions that are typically presumed in multitarget tracking. Two years ago at this conference, we initiated an exploratory analysis of general multitarget pairwise-Markov (MPMC) systems, which weaken the multitarget Markov assumption. Based on this analysis, we derived an exploratory CPHD filter for MPMC systems. Unfortunately, this approach relied on heuristic models in order to incorporate both spatial and cardinality correlation between states and measurements. This paper describes a fully rigorous approach, provided that only cardinality correlation is taken into account. We derive the time-update and measurement-update equations for a CPHD filter describing the evolution of such an MPMC system.

The random infinite set (RFS) approach to information fusion addressed target track-labeling from the outset. The first implementations of RFS filters did not do so because of computational concerns, whereas subsequent implementations employed heuristics. The labeled RFS (LRFS) theory of B.-T. Vo and B.-N. Vo was the first systematic, theoretically rigorous formulation of true multitarget tracking; and led to the generalized labeled multi-Bernoulli (GLMB) filter (the first provably Bayes-optimal multitarget tracking algorithm). This paper addresses the feasibility of theoretically rigorous cardinalized probability hypothesis density (CPHD) filters. We show that an approximation of the GLMB filter, known as the LMB filter, can be reinterpreted as a theoretically rigorous labeled PHD (LPHD) filter. We also prove two characterization theorems for the probability generating functionals (p.g.fl's) of LRFS’s.

In this paper we propose a Superpositional Marginalized δ-GLMB (SMδ-GLMB) filter for multi-target tracking and we provide bootstrap and particle flow particle filter implementations. Particle filter implementations of the marginalized δ-GLMB filter are computationally demanding. As a first contribution we show that for the specific case of superpositional observation models, a reduced complexity update step can be achieved by employing a superpositional change of variables. The resulting SMδ-GLMB filter can be readily implemented using the unscented Kalman filter or particle filtering methods. As a second contribution, we employ particle flow to produce a measurement-driven importance distribution that serves as a proposal in the SMδ-GLMB particle filter. In high-dimensional state systems or for highly- informative observations the generic particle filter often suffers from weight degeneracy or otherwise requires a prohibitively large number of particles. Particle flow avoids particle weight degeneracy by guiding particles to regions where the posterior is significant. Numerical simulations showcase the reduced complexity and improved performance of the bootstrap SMδ-GLMB filter with respect to the bootstrap Mδ-GLMB filter. The particle flow SMδ-GLMB filter further improves the accuracy of track estimates for highly informative measurements.

In a two class label scenario, classi…cation systems may be used to assess whether or not an element of interest belongs to the “target”or “non-target”class. The performance of the system is summarized visually as a trade- o¤ between the proportions of elements correctly labeled as “target”plotted against the proportion of elements incorrectly labeled as “target.” These proportions are empirical estimates of the true and false positive rates, and their trade-o¤ plot is known as a receiver operating characteristic (ROC) curve. Classi…cation performance can be increased, however, if the information provided by multiple systems can be fused together to create a new, combined system. This research focuses on label-fusion as a common method to increase classi…cation performance and quantifying the bias that occurs when misspecifying the partitioning of the underlying event set. This partitioning will be de…ned in terms of what be called within and across label fusion. When incorrect assumptions are made about the partitioning of the event set, bias will occur and both the ROC curve and its optimal parameters will be incorrectly quanti…ed. In this work, we analyze the e¤ects of individual classi…cation system performance, correlation, and target environment on the magnitude of this performance bias. This work will then inspire the development of formulas to adjust optimal performance measures to appropriately re‡ect the fused system performance according to event set partitioning. As such, bias may be appropriately adjusted without redesigning the fused system, allowing greater use of currently fused systems across multiple platforms and environments.

A detection system outputs two distinct labels, thus, there are two errors it can make. The Receiver Operating Characteristic (ROC) function quantifies both of these errors as parameters vary within the system. Combining two detection systems typically yields better performance when a combining rule is chosen appropriately. When detection systems are combined the assumption of independence is usually made in order to simplify the math- ematics, so that we need only combine the individual ROC curve from each system into one ROC curve. This paper investigates label fusion of two detection systems drawn from a single Detection System Family (DSF). Given that one knows the ROC function for the DSF, we seek a formula with the resultant ROC function of the fused detection systems as a function (specifically, a transformation) of the ROC function. In this paper, we derive this transformation for the disjunction and conjunction label rules. Examples are given that demonstrates this transformation. Furthermore, another transformation is given to account for the dependencies between the two systems within the family. Examples will be given that demonstrates these ideas and the corresponding transformation acting on the ROC curve.

We describe a new algorithm for stochastic particle flow filters using Gromov’s method. We derive a simple exact formula for Q in certain special cases. The purpose of using stochastic particle flow is two fold: improve estimation accuracy of the state vector and improve the accuracy of uncertainty quantification. Q is the covariance matrix of the diffusion for particle flow corresponding to Bayes’ rule.

We show the results of numerical experiments for a new algorithm for stochastic particle flow filters designed using Gromov’s method. We derive a simple exact formula for Q in certain special cases. The purpose of using stochastic particle flow is two fold: improve estimation accuracy of the state vector and improve the accuracy of uncertainty quantification. Q is the covariance matrix of the diffusion for particle flow corresponding to Bayes’ rule.

In this paper we discuss two potential areas of intersection between Quantum Information Technologies and Information Fusion. The first area we call Quantum (Data Fusion) and refers to the use of quantum computers to perform data fusion algorithms with classical data generated by quantum and classical sensors. As we discuss, we expect that these quantum fusion algorithms will have a better computational complexity than traditional fusion algorithms. This means that quantum computers could allow the efficient fusion of large data sets for complex multi-target tracking. On the other hand, (Quantum Data) Fusion refers to the fusion of quantum data that is being generated by quantum sensors. The output of the quantum sensors is considered in the form of qubits, and a quantum computer performs data fusion algorithms. Our theoretical models suggest that we expect that these algorithms can increase the sensitivity of the quantum sensor network.

During the 2016 SPIE DSS conference, nine panelists were invited to highlight the trends and opportunities in cyber-physical systems (CPS) and Internet of Things (IoT) with information fusion. The world will be ubiquitously outfitted with many sensors to support our daily living thorough the Internet of Things (IoT), manage infrastructure developments with cyber-physical systems (CPS), as well as provide communication through networked information fusion technology over the internet (NIFTI). This paper summarizes the panel discussions on opportunities of information fusion to the growing trends in CPS and IoT. The summary includes the concepts and areas where information supports these CPS/IoT which includes situation awareness, transportation, and smart grids.

Traditionally, information fusion approaches to meaning making have been integrative or aggregative in nature, creating meaning “containers” in which to put content (e.g., attributes) about object classes. In a large part, this was due to the limits in technology/tools for supporting information fusion (e.g., computers). A different synthesis based approach for meaning making is described which takes advantage of computing advances. The approach is not focused on the events/behaviors being observed/sensed; instead, it is human work centric. The former director of the Defense Intelligence Agency once wrote, “Context is king. Achieving an understanding of what is happening – or will happen – comes from a truly integrated picture of an area, the situation and the various personalities in it…a layered approach over time that builds depth of understanding.”1 The synthesis based meaning making framework enables this understanding. It is holistic (both the sum and the parts, the proverbial forest and the trees), multi-perspective and emulative (as opposed to representational). The two approaches are complementary, with the synthesis based meaning making framework as a wrapper. The integrative approach would be dominant at level 0,1 fusion: data fusion, track formation and the synthesis based meaning making becomes dominant at higher fusion levels (levels 2 and 3), although both may be in play. A synthesis based approach to information fusion is thus well suited for “gray zone” challenges in which there is aggression and ambiguity and which are inherently perspective dependent (e.g., recent events in Ukraine).

Agent-based crowd behaviour consists a significant field of research that has drawn a lot of attention in recent years. Agent-based crowd simulation techniques have been used excessively to forecast the behaviour of larger or smaller crowds in terms of certain given conditions influenced by specific cognition models and behavioural rules and norms, imposed from the beginning. Our research employs conditional event algebra, statistical methodology and agent-based crowd simulation techniques in developing a behavioural econometric model about the selection of certain economic behaviour by a consumer that faces a spectre of potential choices when moving and acting in a multiplex mall. More specifically we try to analyse the influence of demographic, economic, social and cultural factors on the economic behaviour of a certain individual and then we try to link its behaviour with the general behaviour of the crowds of consumers in multiplex malls using agent-based crowd simulation techniques. We then run our model using Generalized Least Squares and Maximum Likelihood methods to come up with the most probable forecast estimations, regarding the agent’s behaviour. Our model is indicative about the formation of consumers’ spectre of choices in multiplex malls under the condition of predefined preferences and can be used as a guide for further research in this area.

The aim of this paper is to present the sensor monitoring and decision level fusion scheme for early fire detection which has been developed in the context of the AF3 Advanced Forest Fire Fighting European FP7 research project, adopted specifically in the OCULUS-Fire control and command system and tested during a firefighting field test in Greece with prescribed real fire, generating early-warning detection alerts and notifications. For this purpose and in order to improve the reliability of the fire detection system, a two-level fusion scheme is developed exploiting a variety of observation solutions from air e.g. UAV infrared cameras, ground e.g. meteorological and atmospheric sensors and ancillary sources e.g. public information channels, citizens smartphone applications and social media. In the first level, a change point detection technique is applied to detect changes in the mean value of each measured parameter by the ground sensors such as temperature, humidity and CO₂ and then the Rate-of-Rise of each changed parameter is calculated. In the second level the fire event Basic Probability Assignment (BPA) function is determined for each ground sensor using Fuzzy-logic theory and then the corresponding mass values are combined in a decision level fusion process using Evidential Reasoning theory to estimate the final fire event probability.

In this paper a solution is presented aiming to assist the early detection and localization of a fire incident by exploiting crowdsourcing and unofficial civilian online reports. It consists of two components: (a) the potential fire incident detection and (b) the visualization component. The first component comprises two modules that run in parallel and aim to collect reports posted on public platforms and conclude to potential fire incident locations. It collects the public reports, distinguishes reports that refer to a potential fire incident and store the corresponding information in a structured way. The second module aggregates all these stored reports and conclude to a probable fire location, based on the amount of reports per area, the time and location of these reports. In further the result is entered to a fusion module which combines it with information collected by sensors if available in order to provide a more accurate fire event detection capability. The visualization component is a fully – operational public information channel which provides accurate and up-to-date information about active and past fires, raises awareness about forest fires and the relevant hazards among citizens. The channel has visualization capabilities for presenting in an efficient way information regarding detected fire incidents fire expansion areas, and relevant information such as detecting sensors and reporting origin. The paper concludes with insight to current CONOPS end user with regards to the inclusion of the proposed solution to the current CONOPS of fire detection.

Particle swarm optimization (PSO) and genetic algorithms (GAs) are two optimization techniques from the field of computational intelligence (CI) for search problems where a direct solution can not easily be obtained. One such problem is finding an optimal set of parameters for the maximally stable extremal region (MSER) algorithm to detect areas of interest in imagery. Specifically, this paper describes the design of a GA and PSO for optimizing MSER parameters to detect stop signs in imagery produced via simulation for use in an autonomous vehicle navigation system. Several additions to the GA and PSO are required to successfully detect stop signs in simulated images. These additions are a primary focus of this paper and include: the identification of an appropriate fitness function, the creation of a variable mutation operator for the GA, an anytime algorithm modification to allow the GA to compute a solution quickly, the addition of an exponential velocity decay function to the PSO, the addition of an ”execution best” omnipresent particle to the PSO, and the addition of an attractive force component to the PSO velocity update equation. Experimentation was performed with the GA using various combinations of selection, crossover, and mutation operators and experimentation was also performed with the PSO using various combinations of neighborhood topologies, swarm sizes, cognitive influence scalars, and social influence scalars. The results of both the GA and PSO optimized parameter sets are presented. This paper details the benefits and drawbacks of each algorithm in terms of detection accuracy, execution speed, and additions required to generate successful problem specific parameter sets.

In this paper, we describe an algorithm and system for optimizing search and detection performance for “items of interest” (IOI) in large-sized images and videos that employ the Rapid Serial Visual Presentation (RSVP) based EEG paradigm and surprise algorithms that incorporate motion processing to determine whether static or video RSVP is used. The system works by first computing a motion surprise map on image sub-regions (chips) of incoming sensor video data and then uses those surprise maps to label the chips as either “static” or “moving”. This information tells the system whether to use a static or video RSVP presentation and decoding algorithm in order to optimize EEG based detection of IOI in each chip. Using this method, we are able to demonstrate classification of a series of image regions from video with an azimuth value of 1, indicating perfect classification, over a range of display frequencies and video speeds.

This paper† describes a technique in which we improve upon the prior performance of the Rapid Serial Visual Presentation (RSVP) EEG paradigm for image classification though the insertion of visual attention distracters and overall sequence reordering based upon the expected ratio of rare to common “events” in the environment and operational context. Inserting distracter images maintains the ratio of common events to rare events at an ideal level, maximizing the rare event detection via P300 EEG response to the RSVP stimuli. The method has two steps: first, we compute the optimal number of distracters needed for an RSVP stimuli based on the desired sequence length and expected number of targets and insert the distracters into the RSVP sequence, and then we reorder the RSVP sequence to maximize P300 detection. We show that by reducing the ratio of target events to nontarget events using this method, we can allow RSVP sequences with more targets without sacrificing area under the ROC curve (azimuth).

Automatic target recognition (ATR) systems, like human photo-interpreters, rely on a variety of visual information for detecting, classifying, and identifying manmade objects in aerial imagery. We describe the integration of a visual learning component into the Image Data Conditioner (IDC) for target/clutter and other visual classification tasks. The component is based on an implementation of a model of the visual cortex developed by Serre, Wolf, and Poggio. Visual learning in an ATR context requires the ability to recognize objects independent of location, scale, and rotation. Our method uses IDC to extract, rotate, and scale image chips at candidate target locations. A bootstrap learning method effectively extends the operation of the classifier beyond the training set and provides a measure of confidence. We show how the classifier can be used to learn other features that are difficult to compute from imagery such as target direction, and to assess the performance of the visual learning process itself.

Vector shoreline (VSL) data is potentially useful in ATR systems that distinguish between objects on land or water. Unfortunately available data such as the NOAA 1:250,000 World Vector Shoreline and NGA Prototype Global Shoreline data cannot be used by themselves to make a land/water determination because of the manner in which the data are compiled. We describe a data fusion approach for creating labeled VSL data using test points from Global 30 Arc-Second Elevation (GTOPO30) data to determine the direction of vector segments; i.e., whether they are in clockwise or counterclockwise order. We show consistently labeled VSL data be used to easily determine whether a point is on land or water using a vector cross product test.

Road Following is a critical component of blindBike, our assistive biking application for the visually impaired. This paper talks about the overall blindBike system and goals prominently featuring Road Following, which is the task of directing the user to follow the right side of the road. This work unlike what is commonly found for self-driving cars does not depend on lane line markings. 2D computer vision techniques are explored to solve the problem of Road Following. Statistical techniques including the use of Gaussian Mixture Models are employed. blindBike is developed as an Android Application and is running on a smartphone device. Other sensors including Gyroscope and GPS are utilized. Both Urban and suburban scenarios are tested and results are given. The success and challenges faced by blindBike’s Road Following module are presented along with future avenues of work.

The solution for Intersection Detection and Crossing to support the development of blindBike an assisted biking system for the visually impaired is discussed. Traffic light detection and intersection crossing are key needs in the task of biking. These problems are tackled through the use of mobile computer vision, in the form of a mobile application on an Android phone. This research builds on previous Traffic Light detection algorithms with a focus on efficiency and compatibility on a resource-limited platform. Light detection is achieved through blob detection algorithms utilizing training data to detect patterns of Red, Green and Yellow in complex real world scenarios where multiple lights may be present. Also, issues of obscurity and scale are addressed. Safe Intersection crossing in blindBike is also discussed. This module takes a conservative “assistive” technology approach. To achieve this blindBike use’s not only the Android device but, an external bike cadence Bluetooth/Ant enabled sensor. Real world testing results are given and future work is discussed.

Recognizing substantially occluded objects in confined spaces is a very challenging problem for ground-based persistent surveillance systems. In this paper, we discuss the ontology inference of occluded object recognition in the context of in-vehicle group activities (IVGA) and describe an approach that we refer to as utilization-based object recognition method. We examine the performance of three types of classifiers tailored for the recognition of objects with partial visibility, namely, (1) Hausdorff Distance classifier, (2) Hamming Network classifier, and (3) Recurrent Neural Network classifier. In order to train these classifiers, we have generated multiple imagery datasets containing a mixture of common objects appearing inside a vehicle with full or partial visibility and occultation. To generate dynamic interactions between multiple people, we model the IVGA scenarios using a virtual simulation environment, in which a number of simulated actors perform a variety of IVGA tasks independently or jointly. This virtual simulation engine produces the much needed imagery datasets for the verification and validation of the efficiency and effectiveness of the selected object recognizers. Finally, we improve the performance of these object recognizers by incorporating human gestural information that differentiates various object utilization or handling methods through the analyses of dynamic human-object interactions (HOI), human-human interactions (HHI), and human-vehicle interactions (HVI) in the context of IVGA.

Partially observable group activities (POGA) occurring in confined spaces are epitomized by their limited observability of the objects and actions involved. In many POGA scenarios, different objects are being used by human operators for the conduct of various operations. In this paper, we describe the ontology of such as POGA in the context of In-Vehicle Group Activity (IVGA) recognition. Initially, we describe the virtue of ontology modeling in the context of IVGA and show how such an ontology and a priori knowledge about the classes of in-vehicle activities can be fused for inference of human actions that consequentially leads to understanding of human activity inside the confined space of a vehicle. In this paper, we treat the problem of “action-object” as a duality problem. We postulate a correlation between observed human actions and the object that is being utilized within those actions, and conversely, if an object being handled is recognized, we may be able to expect a number of actions that are likely to be performed on that object. In this study, we use partially observable human postural sequences to recognition actions. Inspired by convolutional neural networks (CNNs) learning capability, we present an architecture design using a new CNN model to learn “action-object” perception from surveillance videos. In this study, we apply a sequential Deep Hidden Markov Model (DHMM) as a post-processor to CNN to decode realized observations into recognized actions and activities. To generate the needed imagery data set for the training and testing of these new methods, we use the IRIS virtual simulation software to generate high-fidelity and dynamic animated scenarios that depict in-vehicle group activities under different operational contexts. The results of our comparative investigation are discussed and presented in detail.

Multisensor image fusion (e.g. IR with visual) is the process of combining relevant information from two or more images into a single image. The aim is to find an objective quality measure, which can be used in automatic applications, that correlates best with subjective observer trials. Not all combinations of image, algorithm, and test observer can be worked out. In this paper R. Fisher’s Design of Experiments approach based on Latin Squares is used for thinning out the number of experiments for each observer in such observer trials while preserving exactness and reliability of the result.

We present analysis methods that may be used to geolocate emitters using one or more moving receivers. While some of the methods we present may apply to a broader class of signals, our primary interest is locating and tracking ships from short pulsed transmissions, such as the maritime Automatic Identification System (AIS.) The AIS signal is difficult to process and track since the pulse duration is only 25 milliseconds, and the pulses may only be transmitted every six to ten seconds. In this article, we address several problems including accurate TDOA and FDOA estimation methods that do not require searching a two dimensional surface such as the cross-ambiguity surface. As an example, we apply these methods to identify and process AIS pulses from a single emitter, making it possible to geolocate the AIS signal using a single moving receiver.

The receiver operating characteristic (ROC curve), which is a plot of the probability of detection as a function of the probability of false alarm, plays a key role in the classical analysis of detector performance. However, meaningful characterization of the ROC curve is challenging when practically important complications such as variations in source emissions, environmental impacts on the signal propagation, uncertainties in the sensor response, and multiple sources of interference are considered. In this paper, a relatively simple but realistic model for scattered signals is employed to explore how parametric uncertainties impact the ROC curve. In particular, we show that parametric uncertainties in the mean signal and noise power substantially raise the tails of the distributions; since receiver operation with a very low probability of false alarm and a high probability of detection is normally desired, these tails lead to severely degraded performance. Because full a priori knowledge of such parametric uncertainties is rarely available in practice, analyses must typically be based on a finite sample of environmental states, which only partially characterize the range of parameter variations. We show how this effect can lead to misleading assessments of system performance. For the cases considered, approximately 64 or more statistically independent samples of the uncertain parameters are needed to accurately predict the probabilities of detection and false alarm. A connection is also described between selection of suitable distributions for the uncertain parameters, and Bayesian adaptive methods for inferring the parameters.

The paper focuses on extracting scattering centers of radar targets using compressive sensing and using them as features in a target recognition system. It has been shown that a target’s high resolution range profile (HRRP) is sparse in time corresponding to few scatterers that can be associated with target geometry. The recognition system is tested using real radar data of commercial aircraft models. Classification is carried out using distance based and correlation based techniques. Scenarios where the target aspect angle is unknown or known to be within a certain range are also examined.

Small unmanned aerial vehicles (UAV) flying at low altitude are becoming more and more a serious threat in civilian and military scenarios. In recent past, numerous incidents have been reported where small UAV were flying in security areas leading to serious danger to public safety or privacy. The detection and tracking of small UAV is a widely discussed topic. Especially, small UAV flying at low altitude in urban environment or near background structures and the detection of multiple UAV at the same time is challenging. Field trials were carried out to investigate the detection and tracking of multiple UAV flying at low altitude with state of the art detection technologies. Here, we present results which were achieved using a heterogeneous sensor network consisting of acoustic antennas, small frequency modulated continuous wave (FMCW) RADAR systems and optical sensors. While acoustics, RADAR and LiDAR were applied to monitor a wide azimuthal area (360◦) and to simultaneously track multiple UAV, optical sensors were used for sequential identification with a very narrow field of view.

Extracting information from low signal to noise ratio images poses significant challenges. Noise makes extracting spatial features difficult, in particular if extraction of both large, smooth features at the same time as point-like features is required. This work describes a new statistical approach, able to handle both simultaneously, with the capacity of handling both positive and negative contrast signatures. The basic idea in this approach is that each pixel value can represent underlying statistics to a varying degree, depending on how similar it is to samples taken close to it, spatially and/or temporally. If the sample is similar to its surroundings, it is strongly filtered and also affects the filtering of neighboring samples, but if it is significantly different, it will remain largely unfiltered and does not influence neighboring pixel filtering. Simulations show that the filtering maintains energy conservation, significantly limits noise and at the same time maintains signal integrity. The filter is found to adapt to noise characteristics and spatiotemporal variations of the background. The technique is found to be well suited to rocket plume imaging, but is adaptable to a broad range of other applications.

This article discusses the speaker identification for the improvement of the security communication between law enforcement units. The main task of this research was to develop the text-independent speaker identification system which can be used for real-time recognition. This system is designed for identification in the open set. It means that the unknown speaker can be anyone. Communication itself is secured, but we have to check the authorization of the communication parties. We have to decide if the unknown speaker is the authorized for the given action. The calls are recorded by IP telephony server and then these recordings are evaluate using classification If the system evaluates that the speaker is not authorized, it sends a warning message to the administrator. This message can detect, for example a stolen phone or other unusual situation. The administrator then performs the appropriate actions. Our novel proposal system uses multilayer neural network for classification and it consists of three layers (input layer, hidden layer, and output layer). A number of neurons in input layer corresponds with the length of speech features. Output layer then represents classified speakers. Artificial Neural Network classifies speech signal frame by frame, but the final decision is done over the complete record. This rule substantially increases accuracy of the classification. Input data for the neural network are a thirteen Mel-frequency cepstral coefficients, which describe the behavior of the vocal tract. These parameters are the most used for speaker recognition. Parameters for training, testing and validation were extracted from recordings of authorized users. Recording conditions for training data correspond with the real traffic of the system (sampling frequency, bit rate). The main benefit of the research is the system developed for text-independent speaker identification which is applied to secure communication between law enforcement units.

We propose a decision-level architecture that combines synthetic aperture radar (SAR) and an infrared (IR) sensor for automatic target detection. We present a new size-based feature, called target-silhouette to reduce the number of false alarms produced by the conventional target-detection algorithm. Boolean Map Visual Theory is used to combine a pair of SAR and IR images to generate the target-enhanced map. Then basic belief assignment is used to transform this map into a belief map. The detection results of sensors are combined to build the target-silhouette map. We integrate the fusion mass and the target-silhouette map on the decision level to exclude false alarms. The proposed algorithm is evaluated using a SAR and IR synthetic database generated by SE-WORKBENCH simulator, and compared with conventional algorithms. The proposed fusion scheme achieves higher detection rate and lower false alarm rate than the conventional algorithms.

Signal processing techniques are prevalent in a wide range of fields: control, target tracking, telecommunications, robotics, fault detection and diagnosis, and even stock market analysis, to name a few. Although first introduced in the 1950s, the most popular method used for signal processing and state estimation remains the Kalman filter (KF). The KF offers an optimal solution to the estimation problem under strict assumptions. Since this time, a number of other estimation strategies and filters were introduced to overcome robustness issues, such as the smooth variable structure filter (SVSF). In this paper, properties of the SVSF are explored in an effort to detect and diagnosis faults in an electromechanical system. The results are compared with the KF method, and future work is discussed.

SIP is one of the most successful protocols in the field of IP telephony communication. It establishes and manages VoIP calls. As the number of SIP implementation rises, we can expect a higher number of attacks on the communication system in the near future. This work aims at malicious SIP traffic classification. A number of various machine learning algorithms have been developed for attack classification. The paper presents a comparison of current research and the use of classifier fusion method leading to a potential decrease in classification error rate. Use of classifier combination makes a more robust solution without difficulties that may affect single algorithms. Different voting schemes, combination rules, and classifiers are discussed to improve the overall performance. All classifiers have been trained on real malicious traffic. The concept of traffic monitoring depends on the network of honeypot nodes. These honeypots run in several networks spread in different locations. Separation of honeypots allows us to gain an independent and trustworthy attack information.

The sequence generator generates a sequence of pseudorandom binary numbers using a linear-feedback shift register (LFSR). This block implements LFSR using a simple shift register generator (SSRG, or Fibonacci) configuration. In this article we introduce the concept of probabilistic binary element provides requirements, which ensure compliance with the criterion of "uniformity" in the implementation of the basic physical generators uniformly distributed random number sequences. Based on these studies, we obtained an analytic relation between the parameters of the binary sequence and parameters of a numerical sequence with the shift register output. The received analytical dependencies can help in evaluating the statistical characteristics of the processes in solving problems of statistical modeling. It is supposed that the formation of the binary sequence output from the binary probabilistic element is produced using a physical noise process. It is shown that the observed errors in statistical modeling using pseudo-random numbers do not occur if the model examines linear systems with constant parameters, but in case models of nonlinear systems, higher order moments can have a Gaussian distribution.

The probabilistic multiple-hypothesis tracker (PMHT), a tracking algorithm of considerable theoretical elegance based on the expectation-maximization (EM) algorithm, will be considered for the problem of multiple target tracking (MTT) with multiple sensors in clutter. Aside from position observations, continuous measurements associated with the unique and constant feature of each target are incorporated to jointly estimate the states and feature of the targets for the sake of tracking and classification, leading to a bootstrapped implementation of the PMHT. In addition, we rederived the information matrix for the big state vector stacking states for all the targets at all the time steps during the observation time. Simulation results have been conducted for both closely spaced and well separated scenarios with and without feature measurements. The normalized estimation error squared (NEES) calculated using the information matrix for both scenarios with and without feature measurements are within the 95% probability region. In other words, the estimates are consistent with the corresponding covariances.

This paper describes the ways how to measure the airburst height of projectiles and rockets. In general, the airburst height could be determined by using triangulation method or the images from the camera installed on the radar. There are some limitations in these previous methods when the missiles impact the sea surface. To apply triangulation method, the cameras should be installed so that the lines of sight intersect at angles from 60 to 120 degrees. There could be no effective observation towers to install the optical system. In case the range of the missile is more than 50km, the images from the camera of the radar could be useless. This paper proposes the method to measure the airburst height of sea impact projectile by using a single camera. The camera would be installed on the island near to the impact area and the distance could be computed by using the position and attitude of camera and sea level. To demonstrate the proposed method, the results from the proposed method are compared with that from the previous method.

In this paper, improving data association process by increasing the probability of detecting valid data points (measurements obtained from ESM/RADAR system) in the presence of noise for location and target tracking are discussed. This develop a multisensor data association algorithm that fuses information from the multiple ESM receiver and surveillance RADAR. The develop a novel algorithm by self-organizing fuzzy neural network (SO-FNN) for multiple ESM-to-ESM (measurement-to-measurement data association) and ESMs-to-RADAR (track-to-track data association) problem in dense clutter environment. An adaptive search based on SO-FNN of the distance threshold measure is then used to detect valid filtered data point for data association. Simulation results demonstrate the effectiveness and better performance when compared to conventional algorithm. The paper is organized as follows. Section 1 is the problem formulation. Section 2 design the new data association algorithm based on SO-FNN data association system design. Section 3 describes ESM-to-ESM and ESM-to-RADAR (measurement-to-measurement data association and track-to-track data association) scenario and the simulation results are presented and discussed. The summary are drawn in section 4, respectively.

Electromagnetic signals and the requirements of timely response have been a rapid growth in modern electronic warfare. Although jammers are limited resources, it is possible to achieve the best electronic warfare efficiency by tactical decisions. This paper proposes the intelligent electronic warfare decision support system. In this work, we develop a novel hybrid algorithm, Digital Pheromone Particle Swarm Optimization, based on Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO) and Shuffled Frog Leaping Algorithm (SFLA). We use PSO to solve the problem and combine the concept of pheromones in ACO to accumulate more useful information in spatial solving process and speed up finding the optimal solution. The proposed algorithm finds the optimal solution in reasonable computation time by using the method of matrix conversion in SFLA. The results indicated that jammer allocation was more effective. The system based on the hybrid algorithm provides electronic warfare commanders with critical information to assist commanders in effectively managing the complex electromagnetic battlefield.