In this paper, a new joint target tracking and classification technique based on Observable Operator Models (OOM) is considered. The OOM approach, which has been proposed as a better alternative to the Hidden Markov Model (HMM), is used to model the stochastic process of target classification. These OOMs afford both mathematical simplicity and algorithmic efficiency compared to HMM. Conventional classification techniques use only the feature information from target signatures. The proposed OOM based classification technique incorporates the target-to-sensor orientation together with a sequence of feature information from multiple sensors. The target-to-sensor orientation evolves over time and the target aspect is important in determining the target classes. The multi-aspect classification is modeled using OOM to handle unknown target orientation. This algorithm exploits the inter-dependency of target state and the target class, which improves both the state estimates and classification of each target. Measurement ambiguity is present in both kinematic and feature measurement and therefore, the OOM based classifier is integrated into the multiframe data association framework that is used to resolve measurement origin uncertainties. This technique enables one to overcome ambiguity in feature measurements while improving track purity. A two dimensional example demonstrates the merits of the proposed OOM based joint target tracking and classification algorithm.

Part of target tracking is the use of existing information to predict behavior. Often, this part of the tracking algorithm is used for data association and the update step of the fusion process as well as, recently, to guide weapons to the target. Another application is to estimate the point of impact of enemy ballistic munitions. The determination of the actual threat posed by enemy position permits prioritization of targets for response and can limit the need to expose friendly units. The flight trajectory of ballistic ordinance, while theoretically understood, can be affected by a number of unknown factors, such as air turbulence and drag. To accurately predict the projectile's flight path, an adaptive Kalman filter approach is proposed to track and predict the target trajectory. The Kalman filter uses a neural network as an on-line function approximator to improve the motion model of the target tracking. The neural network is trained in conjunction with the tracking algorithm. Both track states and neural network weights use the same residual information. The neural network model is added to the motion model and provides an improved track prediction of a mortar impact time and its location. Analysis of the approach's performance is compared to varying tracking systems with different levels of target motion accuracy. Two a priori motion models for the neural approach are applied: a straight-line motion model and a basic ballistic trajectory. The results show the ability of the technique to provide accurate target predictions with limited a priori information.

A new non-linear filtering and predication (NFP) algorithm with input estimation is proposed for maneuvering target tracking. In the proposed method, the acceleration level is determined by a decision process, where a least squares (LS) estimator plays a major role to detect target maneuvering within a sliding window. In this paper, we first illustrate that the optimal solution to minimize the mean squared error (MSE) must consider a trade-off between the bias and error variance. For the application of target tracking, we then derive the MSE of target positions in a close form by using orthogonal space decompositions. Then we discuss the NFP estimator, and evaluate how well the approach potentially works in the case of given system parameters. Comparing with the traditional unbiased minimum variance filter (UMVF), Kalman filter, and interactive multiple model (IMM) algorithms, numerical results show that the newly proposed NFP method performs comparable or better in all scenarios with less computational requirements.

It is critical in urban environments to not only track cars and tanks; but also individuals. Tracking dismounts, whereby an individual exits a car, can be done using conventional Electro-Optical (full color) or Infrared (thermal) cameras. However, EO/IR systems are subject to weather and line-of-sight conditions (i.e. person blocked by cloud) as well are degraded for long ranges. In this study, we pursue the use of radar images for dismount tracking. Radar dismount tracking will not entail the same robust features for person identification as EO systems; however, by being able to maintain track in all-weather conditions would afford friendly forces a location of all moving individuals. We show, using a feature-based tracker, that dismount detection, tracking, and potential intent, is possible. Radio Frequency (RF) tracking of dismounts is a relatively new concept because the data has not been available. By forming a data set based on the POSER^TM program, and post-processing the data, we are interested in exploring the possibility of RF dismount tracking. In this paper, we (1) explore the generation of RF dismount data, (2) apply feature-based tracking algorithm to locate the moving target, and (3) assess the performance.

Tracking multiple surface targets with a single steerable airborne video sensor is accomplished by several interrelated functions: (i) image registration for camera motion compensation and accurate image-to-ground mapping, (ii) video processing for object detection and feature extraction, (iii) target tracking for detection association and track creation and maintenance, (iv) signature extraction and exploitation, and (v) sensor resource management for the generation of sensor steering commands. The first function is often overlooked, but has a significant impact in the performance of the overall system. A rudimentary registration can be achieved by using the platform location and attitude as well as the sensor orientation and field of view information, but the accuracy of this registration is typically poor due to inertial navigation system errors, particularly in small unmanned aerial vehicles with cost and hardware limitations. Successful registration of successive frames enables the use of multiple frame video processing for improved object detection and provides stable image-to-ground mapping for improved data association by the tracker. In systems with a steerable sensor that slews back and forth to track more than one target simultaneously, the image registration module creates and maintains multiple mosaics corresponding to the different tracking areas. In this paper we discuss primarily the image registration module and the system of coordinate frames that is maintained to improve data association and tracking.

This paper proposes a global motion model estimation and target detection algorithm for surveillance and tracking applications. The proposed algorithm analyzes the foreground-background structure of a video frame, and detects objects with independent motions. Each video frame is first segmented into regions where image intensity and motion fields are homogeneous. Then global motion model fitting is accomplished using linear regression of motion vectors through iterations of region search. With the use of non-parametric estimation of motion field, the proposed methods is more efficient than direct estimation of motion parameter; and it is able to detect outliers where independent moving targets are located. The proposed algorithm is more computationally efficient than parametric motion estimation, and also more robust than a variety of background compensation based detection.

In this paper, a localized particle subset method is proposed to solve target tracking problem in a Bayesian inference framework. Instead of using all particles to estimated the posterior probability density function (pdf) of targets, a subset is used. This subset of particles is selected by estimated motion of the targets. The weights of particles are updated by the 3D Hausdroff distances between target appearance model and samples. The proposed method is highly efficient in utilizing the particles, which consequently results in reduction of samples utilized in the prediction and update processes. It is also able to alleviate the sample degeneracy and impoverishment problems in the sampling process. Experiments show that the computation complexity for localized particle subset tracker is reduce to a fraction of that of the Sequential Importance (SIS) tracker but with compatible performance.

In this paper we propose a cooperative control algorithm for a group of UAVs carrying out surveillance --- search, tracking and classification --- over a large region which includes a number of targets. The goal is to track and classify detected targets as well as search for new targets. The UAVs are assumed to be equipped with Ground Moving Target Indicator (GMTI) radars, which measure the locations of moving ground targets as well as their radial velocities (Doppler). In addition, a classification sensor is mounted on each UAV that can obtain target class information. The surveillance region is divided into a number of sectors and it is assumed that the GMTI sensor on each UAV scans a fixed number of such sectors in each period of its operation. The sensor responsible for class information can scan only a small circular region around the predicted position of a target. In this paper, a decentralized cooperative control algorithm is proposed, according to which each UAV transmits the current scan information (either kinematic or class information) and detection information (including "negative information") to the other UAVs. Each UAV makes its scan decision and path decision separately, based on information-based objective functions, which incorporate target state information as well as target detection probability and survival probability due to possible hostile fire by targets and collision with other UAVs. The proposed algorithm requires limited communication and modest computation and it can handle failure in communication and loss of UAVs.

The use of measures from information theory to evaluate the expected utility of a set of candidate actions is a popular method for performing sensor resource management. Shannon entropy is a standard metric for information. Past researchers have shown^1-5 that the discrete entropy formula can measure the quality of identification information on a target, while the continuous entropy formula can measure kinematic state information of a target. In both cases, choosing controls to minimize an objective function proportional to entropy will improve ones information about the target. However, minimizing entropy does not naturally promote detection of new targets or "wide area surveillance" (WAS). This paper outlines a way to use Shannon entropy to motivate sensors to track (partially) discovered targets and survey the search space to discover new targets simultaneously. Results from the algorithmic implementation of this method show WAS being favored when most targets in the search space are undiscovered, and tracking of discovered targets being favored when most targets are in track. The tradeoff between these two competing objectives is adjusted by the objective function automatically and dynamically.

The goal of sensor resource management (SRM) is to allocate resources appropriately in order to gain as much information as possible about a system. We introduce a centralized non-myopic planning algorithm, C-SPLAN, that uses sparse sampling to estimate the value of resource assignments. Sparse sampling is related to Monte Carlo simulation. In the SRM problem we consider, our network of sensors observes a set of tracks; each sensor can be set to operate in one of several modes and/or viewing geometries. Each mode incurs a different cost and provides different information about the tracks. Each track has a kinematic state and is of a certain class; the sensors can observe either or both of these, depending on their mode of operation. The goal in this problem is to maximize the overall rate of information gain, i.e. rate of improvement in kinematic tracking and classification accuracy of all tracks in the Area of Interest. The rate is defined by several metrics with the cost of the sensor mode being the primary factor. We compare C-SPLAN's performance on several tracking and target identification problems to that of other algorithms.

Valued situation information is introduced as a measure which can be used to determine which information to obtain with a heterogeneous sensor system in a non-stationary, dynamic environment. Situation information is based on the change in the total Kullback-Leibler divergence of a particular type of influence diagram which we call a situation information expected value network (SIEV Net). The SIEV Net weights situation chance nodes by their dynamic mission goal values as expressed in the topmost goals of a mission goal lattice and computes the valued expected information which will result if a particular information request is selected. The sensor information request which yields the largest valued expected information is then instantiated. The result is the greatest reduction in uncertainty of our situation estimate while constrained to maximize the mission value of the information obtained.

A recently developed fuzzy logic resource allocation algorithm that enables a collection of unmanned air vehicles (UAVs) to automatically cooperate as they make meteorological measurements will be discussed. The goal of the UAVs' coordinated effort is to measure the atmospheric index of refraction. Once in flight no human intervention is required. A fuzzy logic based planning algorithm determines the optimal trajectory and points each UAV will sample, while taking into account the UAVs' risk, reliability, and mission priority for sampling in certain regions. It also considers fuel limitations, mission cost, and related uncertainties. The real-time fuzzy control algorithm running on each UAV renders the UAVs autonomous allowing them to change course immediately without consulting with any commander, requests other UAVs to help, changes the points that will be sampled when observing interesting phenomena, or to terminate the mission and return to base. The control algorithm allows three types of cooperation between UAVs. The underlying optimization procedures including the fuzzy logic based cost function, the fuzzy logic decision rule for UAV path assignment, the fuzzy algorithm that determines when a UAV should alter its mission to help another UAV and the underlying approach to quantifying risks are discussed. Significant simulation results will show the planning algorithm's effectiveness in initially selecting UAVs and determining UAV routes. Likewise, simulation shows the ability of the control algorithm to allow UAVs to effectively cooperate to increase the UAV team's likelihood of success.

We consider the sensor management problem arising in air-to-ground tracking of moving targets. The sensing-tracking system includes a radar and a feature-aided tracker. The radar collects target-signature data in high-resolution-radar (HRR) mode. The tracker is using the collected HRR-signature data to create and maintain target-track identification information. More specifically, the tracker is learning target-track profiles from the collected signature data, and is using these profiles to resolve the potential report-to-track or track-to-track association ambiguities. In this paper, we focus on the management of the HRR-signature data collection. Specifically, the sensor management problem is to determine where to collect signature data on targets in time so as to optimize the utility of the collected data. As with other sensor management problems, determining the optimal data collection is a hard combinatorial problem due to many factors including the large number of possible sensor actions and the complexity of the dynamics. The complexity of the dynamics stems in part from the presence of the sensor slew time. A distinguishing feature of the sensor management problem considered here is that the HRR-signature data collected during the learning phase has no immediate value. To optimize the data collections, a sensor manager must look sufficiently far into the future to adequately trade-off alternative plans. Here, we propose some farsighted algorithms, and evaluate them against a sequential scanning and a greedy algorithm. We present our simulation results obtained by applying these algorithms to a problem of managing a single sensor providing HRR-signature data.

This paper presents a novel, scalable optimization approach to coordinating multiple sensor resources to track and discriminate targets. The sensor resource management problem is one of allocating sensors on a time scale of seconds in a closed loop that includes a target tracker and discrimination system. Allocations specify how the sensors should be pointed and the modes in which they should operate. The objective is to optimize the collection of data to meet tracking and discrimination goals. Quality of the data collected will be different for different allocations, in part, because the quality of the data collected on a target by a sensor depends on the relative geometry between the sensor and target. The optimization of the data collection is to be done subject to constraints on sensors' fields of view as well as time required to slew a sensor into position, if it has mechanical gimbals, and to collect data in the selected mode. The problem is challenging because of the large number of possible sensor allocations as well as the complex dynamics. For this problem, we have developed a novel, approximate dynamic programming algorithm, a type of rollout algorithm, to optimize sensor allocations to coordinate multiple sensor resources. The approach is scalable to realistically sized problems. The paper overviews the approach and results from applying the algorithm.

Modern sensors are able to rapidly change mode of operation and steer between physically separated objects. While control of such sensors over a rolling planning horizon can be formulated as a dynamic program, the optimal solution is inevitably intractable. In this paper, we consider the control problem under a restricted family of policies and show that the essential sensor control trade-offs are still captured. The advantage of this approach is that one can obtain the optimal policy within the restricted class in a tractable fashion, in this case by using the auction algorithm. The approach is well-suited for problems in which a single sensor (or group of sensors) is being used to track many targets using a heterogeneous sensor model, i.e., where the quality of observations varies with object state, such as due to obscuration. Our algorithm efficiently weighs the rewards achievable by observing each target at each time to find the best sensor plan within the restricted set. We extend this approach using a roll-out algorithm, to handle additional cases such as when observations take different amounts of time to complete.

A prototype image processing system has recently been developed which generates, displays and analyzes threedimensional ladar data in real time. It is based upon a suite of novel algorithms that transform raw ladar data into cleaned 3D images. These algorithms perform noise reduction, ground plane identification, detector response deconvolution and illumination pattern renormalization. The system also discriminates static from dynamic objects in a scene. In order to achieve real-time throughput, we have parallelized these algorithms on a Linux cluster. We demonstrate that multiprocessor software plus Blade hardware result in a compact, real-time imagery generation adjunct to an operating ladar. Finally, we discuss several directions for future work, including automatic recognition of moving people, real-time reconnaissance from mobile platforms, and fusion of ladar plus video imagery. Such enhancements of our prototype imaging system can lead to multiple military and civilian applications of national importance.

A major challenge for ATR evaluation is developing an accurate image truth that can be compared to an ATR algorithm's decisions to assess performance. We have developed a semi-automated video truthing application, called START, that greatly improves the productivity of an operator truthing video sequences. The user, after previewing the video selects a set of salient frames (called "keyframes"), each corresponding to significant events in the video. These keyframes are then manually truthed. We provide a spectrum of truthing tools that generates truth for additional frames from the keyframes. These tools include: fully-automatic feature tracking, interpolation, and completely manual methods. The application uses a set of diagnostic measures to manage the user's attention, flagging portions in the video for which the computed truth needs review. This changes the role of the operator from raw data entry, to that of expert appraiser supervising the quality of the image truth. We have implemented a number of graphical displays summarizing the video truthing at various timescales. Additionally, we can view the track information, showing only the lifespan information of the entities involved. A combination of these displays allows the user to manage their resources more effectively. Two studies have been conducted that have shown the utility of START: one focusing on the accuracy of the automated truthing process, and the other focusing on usability issues of the application by a set of expert users.

Future military helicopters will be provided with multiple information sources for self-protection and reconnaissance, e.g. imaging IR, laser radar and GPS. In addition, knowledge bases like maps, aerial images, geographical information (GIS) and other previously acquired data can be used for the interpretation of the current scenario. To support the mission, results of data fusion and information management have to be presented to the pilot in an appropriate way. This paper describes concepts and results of our work on IR and laser data fusion for airborne systems. Data is gathered by forward-looking sensors mounted in a helicopter. For further improvement, fusion with collateral information (laser elevation data, aerial images) is used for change detection and definition of regions of interest with respect to the stored and continuously updated database. Results are demonstrated by the analysis of an exemplary data set, showing a scenario with a group of vehicles. Two moving vehicles are detected automatically in both channels (IR, laser) and the results are combined to achieve improved visualization for the pilot.

The multitarget recursive Bayes nonlinear filter is the theoretically optimal approach to multisensor-multitarget detection, tracking, and identification. For applications in which this filter is appropriate, it is likely to be tractable for only a small number of targets. In earlier papers we derived closed-form equations for an approximation of this filter based on propagation of a first-order multitarget moment called the probability hypothesis density (PHD). In a recent paper, Erdinc, Willett, and Bar-Shalom argued for the need for a PHD-type filter which remains first-order in the states of individual targets, but which is higher-order in target number. In this paper we show that this and much more is possible. We derive a closed-form cardinalized PHD (CPHD), filter, which propagates not only the PHD but also the entire probability distribution on target number.

This paper presents a Sequential Monte Carlo (SMC) Probability Hypothesis Density (PHD) algorithm for decentralized state estimation from multiple platforms. The proposed algorithm addresses the problem of communicating and fusing track information from a team of multiple sensing platforms detecting and tracking multiple targets in the surveillance region. Each sensing platform makes multiple, noisy measurements of an underlying, time-varying state that describes the monitored system. The monitored system involves potentially nonlinear target dynamics described by Markovian state-space model, nonlinear measurements, and non-Gaussian process and measurement noises. Each sensing platform reports measurements to a node in the network, which performs sequential estimation of the current system state using the probability hypothesis density (PHD) filter, which propagates only the first-order statistical moment of the full target posterior of the multi-target state. A sequential Monte Carlo method is used to implement the filter. The crucial consideration is what information needs to be transmitted over the network in order to perform online estimation of the current state of the monitored system, whilst attempting to minimize communication overhead. Simulation results demonstrate the efficiency of the proposed algorithm for a team of bearing only sensors.

In last year's conference we demonstrated new results using a foundational, joint control-theoretic approach to situation assessment (SA) and SA sensor management that is based on a "dynamic situational significance map", the maximization of the expected number of targets of tactical interest, and approximate multitarget filters (specifically, first-order multitarget moment filters and multi-hypothesis correlator (MHC) engines). This year we report on the following new developments and extensions: (1) a tactical significance function based on the fusion of different ambiguous attributes from several different sources; (2) a Bayes' belief network formulation for multi-target tracking and information fusion; and (3) a recursive closed form expression for the posterior expected number of targets of interests (PENTIs) for ANY number of sources. Results of testing this sensor management algorithm with significance maps defined in terms of targets/attributes interrelationships using simplified battlefield situations demonstrate that these new advancements allow for a better SA, and a more efficient SA sensor management.

Sensor management in support of Level 1 data fusion (multisensor integration), or Level 2 data fusion (situation assessment) requires a computationally tractable multitarget filter. The theoretically optimal approach to this multi-target filtering is a suitable generalization of the recursive Bayes nonlinear filter. However, this optimal filter is intractable and computationally challenging that it must usually be approximated. We report on the approximation of a multi-target non-linear filtering for Sensor Management that is based on the particle filter implementation of Stein-Winter probability hypothesis densities (PHDs). Our main focus is on the operational utility of the implementation, and its computational efficiency and robustness for sensor management applications. We present a multitarget Particle Filter (PF) implementation of the PHD that include clustering, regularization, and computational efficiency. We present some open problems, and suggest future developments. Sensor management demonstrations using a simulated multi-target scenario are presented.

Much effort has been expended on devising "conversions" of one uncertainty representation scheme to another- fuzzy to probabilistic, Dempster-Shafer to probabilistic, to fuzzy, etc. Such efforts have been hindered by the fact that uncertainty representation formalisms vary considerably in the degree of complexity of information which they encode. For example, 2^M - 1 numbers are required to specify a Dempster-Shafer basic mass assignment (b.m.a.) on a space with M elements; whereas only M - 1 numbers are required to specify a probability distribution on the same space. Consequently, any conversion of b.m.a.'s to probability distributions will result in a huge loss of information. In addition, conversion from one uncertainty representation formalism to another should be consistent with the data fusion methodologies intrinsic to these formalisms. For b.m.a.'s to be consistently converted to fuzzy membership functions, for example, Dempster's combination should be transformed into fuzzy conjunction in some sense. In this paper we show that a path out of such quandaries is to realize that in many applications all information must ultimately be reduced to state estimates and covariances. Adopting a Bayesian approach, we identify Bayes-invariant conversions between various uncertainty representation formalisms.

Tracking people and vehicles in an urban environment using video cameras onboard unmanned aerial vehicles has drawn a great deal of interest in recent years due to their low cost compared with expensive radar systems. Video cameras onboard a number of small UAVs can provide inexpensive, effective, and highly flexible airborne intelligence, surveillance and reconnaissance as well as situational awareness functions. The perspective transformation is a commonly used general measurement model for the video camera when the variation in terrain height in the object scene is not negligible and the distance between the camera and the scene is not large. The perspective transformation is a nonlinear function of the object position. Most video tracking applications use a nearly constant velocity model (NCVM) of the target in the local horizontal plane. The filtering problem is nonlinear due to nonlinearity in the measurement model. In this paper, we present algorithms for quantifying the degree of nonlinearity (DoN) by calculating the differential geometry based parameter-effects curvature and intrinsic curvature measures of nonlinearity for the video tracking problem. We use the constant velocity model (CVM) of a target in 2D with simulated video measurements in the image plane. We have presented preliminary results using 200 Monte Carlo simulations and future work will focus on detailed numerical results. Our results for the chosen video tracking problem indicate that the DoN is low and therefore, we expect the extended Kalman filter to be reasonable choice.

In applications in which even the best EKFs and MHTs may perform poorly, the single-target and multi-target Bayes nonlinear filters become potentially important. In recent years, new implementation techniques such as sequential Monte Carlo (a.k.a. particle-system) have emerged that, when hosted on ever more inexpensive, smaller, and powerful computers, make these filters potentially computationally tractable for real-time applications. A methodology for preliminary test and evaluation (PT&E) of the relative strengths and weaknesses of these algorithms is becoming increasingly necessary. The purpose of PT&E is to (1) assess the broad strengths and weaknesses of various algorithms or algorithm types; (2) justify further algorithm development; and (3) provide guidance as to which algorithms are potentially useful for which applications. At last year's conference we described our plans for the development of a PT&E tool, MENTAT. In this paper we report on current progress. Our implementation is MATLAB-based, and harnesses the GUI-building capabilities of the well-known MATLAB package, SIMULINK.

The Gaussian Mixure Probability Hypothesis Density (GM-PHD) Multi-target Tracker was developed as an extension to the GM-PHD filter to provide track continuity. The algorithm is demonstrated on forward-looking sonar data with clutter and is compared with the results from the Particle PHD filter.

Theoretically, fusion of measurements is optimal, and fusion of tracks is suboptimal. However, in the real world it frequently turns out that track fusion is superior to measurement fusion. The purpose of this paper is to explain the main reasons for this and to describe a measurement fusion algorithm that is robust in the real world. The two major reasons that measurement fusion fails to perform as well as expected in the real world are: (1) limited sensor resolution, and (2) residual sensor bias errors. Designing robust multi-sensor data fusion algorithms that work well in the real world requires understanding the physics of both sensor resolution and sensor bias errors.

The concept of network centric operations is a process the military is just beginning to struggle with. Distributed information and data fusion provides the foundation to determine target track state estimation within a network centric structure. This paper addresses the overall effectiveness of two popular fusion approaches - the information filter and the channel filter for selected network architectures. Specifically, we compare algorithm performance for two special network cases - the cyclic network and the broadcast network. We will show the scalability of the channel filter when agents are added to the cyclic network and provide an assessment under conditions of random link failure between agents.

Real-time fusion of data collected from a variety of radars that acquire information from multiple perspectives and/or different frequencies, is being shown to provide a more accurate picture of the adversary threat cloud than any single radar or group of radars operating independently. This paper describes a cooperative multi-sensor approach in which multiple radars operate together in a non-interference limited manner, and where decision algorithms are applied to optimize the acquisition, tracking, and discrimination of moving targets with low false alarm rate. The approach is twofold: (i) measure and process radar returns in a shared manner for target feature extraction by exploiting frequency and spatial diversity; and (ii) employ feature-aided track/fusion algorithms to detect, discriminate, and track real targets from the adversary noise cloud. The results of computer simulations are provided that demonstrate the advantages of this approach.

Significant advances in the performance of ATR systems can be made when fusing individual classification systems into a single combined classification system. Often, these individual systems are dependent, or correlated, with one another. Additionally, these systems typically assume that two outcome labels, (for instance "target" and "non-target") exist. Little is known about the performance of fused classification systems when multiple outcome labels are used. In this paper, we propose a methodology for quantifying the performance of the fused classifier system using multiple labels. Specifically, a performance measure for a fused classification system using two classifiers and multiple labels will be developed. The performance measure developed is based on the Receiver Operating Characteristic (ROC) curve. The ROC curve in a two-label system has been well defined and used extensively, in not only ATR applications, but also other engineering and biomedical applications. A ROC manifold is defined and use in order to incorporate the multiple labels. An example of this performance measure for a given fusion rule and multiple labels is given.

Automatic Target Recognition (ATR) system's performance is quantified using Receiver Operating Characteristic (ROC) curves (or ROC manifolds for more than two labels) and typically the prior probabilities of each labeled-event occurring. In real-world problems, one does not know the prior probabilities and they have to be approximated or guessed, but usually one knows their range or distribution. We derive an objective functional that quantifies the robustness of an ATR system given: (1) a set of prior probabilities, and (2) a distribution of a set of prior probabilities. The ATR system may have two labels or more. We demonstrate the utility of this objective functional with examples, and show how it can be used to determine the optimal ATR system from a family of systems.

A model-based multi-sensor fusion framework has previously been developed that supports improved target recognition by fusing target signature information obtained from sensor imagery [1], [2]. Image- based signature features, however, are not the only source of information that may be exploited to advantage by a target recognition system. This paper presents a review of the key features of the model-based fusion framework and shows how it can be expanded to support information derived from imaging sensors as well as data from other non-imaging sources. The expanded model-based multi-source framework supports not only the combination of image data, such as Synthetic Aperture Radar (SAR) and electro-optical (EO), but also various types of non-image data that may be derived from those, or other sensor measurements. The paper illustrates the flexibility of the model-based framework by describing the combination of spatial information from an imaging sensor with scattering characteristics derived from polarimetric phase history data. The multi-source fusion is achieved by relating signature features to specific structural elements on the 3-D target geometry. The 3-D model is used as a sensor neutral, view independent, common reference for the combination of multi-source information.

Image fusion techniques have been used for variety of applications like medical imaging, navigation, homeland security and most importantly in military requirements. Different techniques for image fusion are there and already being extended for real time video fusion. In this paper, a new technique for video image fusion has been given. We exploit fuzzy techniques for image fusion. This approach has already been implemented for multi image fusion for different applications. In the fuzzy approach, pixel of one image is fused with the corresponding pixel value of other image. Fusion is based on the associated rules and membership grades of the frames. For the video image fusion, frames are extracted from the two incoming videos and registered. Size and distortion of the frames are checked for the suitability of the fusion process. After frame wise fusion using fuzzy approach, they are sequenced back for video display. Various other issues like real time implementation, scene effect, adaptation required according to application and image alignments have been discussed. We hope that algorithm developed for video image fusion process in this paper will prove to be very effective for real time image sensor fusion process.

We address a new approach to the problem of improvement of the quality of multi-grade spatial-spectral images provided by several remote sensing (RS) systems as required for environmental resource management with the use of multisource RS data. The problem of multi-spectral reconstructive imaging with multisource information fusion is stated and treated as an aggregated ill-conditioned inverse problem of reconstruction of a high-resolution image from the data provided by several sensor systems that employ the same or different image formation methods. The proposed fusionoptimization technique aggregates the experiment design regularization paradigm with neural-network-based implementation of the multisource information fusion method. The maximum entropy (ME) requirement and projection regularization constraints are posed as prior knowledge for fused reconstruction and the experiment-design regularization methodology is applied to perform the optimization of multisource information fusion. Computationally, the reconstruction and fusion are accomplished via minimization of the energy function of the proposed modified multistate Hopfield-type neural network (NN) that integrates the model parameters of all systems incorporating a priori information, aggregate multisource measurements and calibration data. The developed theory proves that the designed maximum entropy neural network (MENN) is able to solve the multisource fusion tasks without substantial complication of its computational structure independent on the number of systems to be fused. For each particular case, only the proper adjustment of the MENN's parameters (i.e. interconnection strengths and bias inputs) should be accomplished. Simulation examples are presented to illustrate the good overall performance of the fused reconstruction achieved with the developed MENN algorithm applied to the real-world multi-spectral environmental imagery.

Situation assessment (SA) involves deriving relations among entities, e.g., the aggregation of object states (i.e. classification and location). While SA has been recognized in the information fusion and human factors literature, there still exist open questions regarding knowledge representation and reasoning methods to afford SA. For instance, while lots of data is collected over a region of interest, how does this information get presented to an attention constrained user? The information overload can deteriorate cognitive reasoning so a pragmatic solution to knowledge representation is needed for effective and efficient situation understanding. In this paper, we present issues associated with Level 2 (Situation Assessment) including: (1) user perception and perceptual reasoning representation, (2) knowledge discovery process models, (3) procedural versus logical reasoning about relationships, (4) user-fusion interaction through performance metrics, and (5) syntactic and semantic representations. While a definitive conclusion is not the aim of the paper, many critical issues are proposed in order to characterize future successful strategies to knowledge representation and reasoning strategies for situation assessment.

While great progress has been made in the lowest levels of data fusion, practical advances in behavior modeling and prediction remain elusive. The most critical limitation of existing approaches is their inability to support the required knowledge modeling and continuing refinement under realistic constraints (e.g., few historic exemplars, the lack of knowledge engineering support, and the need for rapid system deployment). This paper reports on our ongoing efforts to develop Propheteer, a system which will address these shortcomings through two primary techniques. First, with Propheteer we abandon the typical consensus-driven modeling approaches that involve infrequent group decision making sessions in favor of an approach that solicits asynchronous knowledge contributions (in the form of alternative future scenarios and indicators) without burdening the user with endless certainty or probability estimates. Second, we enable knowledge contributions by personnel beyond the typical core decision making group, thereby casting light on blind spots, mitigating human biases, and helping maintain the currency of the developed behavior models. We conclude with a discussion of the many lessons learned in the development of our prototype Propheteer system.

Modern combat aircraft pilots increasingly rely on high-level fusion models (JDL Levels 2/3) to provide real-time engagement support in hostile situations. These models provide both Situational Awareness (SA) and Threat Assessment (TA) based on data and the relationships between the data. This information represents two distinct classes of uncertainty: vagueness and ambiguity. To address the needs associated with modeling both of these types of data uncertainty, an innovative hybrid approach was recently introduced, combining probability theory and possibility theory into a unified computational framework. The goal of this research is to qualitatively and quantitatively address the advantages and disadvantages of adopting this hybrid framework as well as identifying instances in which the combined model outperforms or is more appropriate than more classical inference approaches. To accomplish this task, domain specific models will be developed using different theoretical approaches and conventions, and then evaluated in comparison to situational ground truth to determine their accuracy and fidelity. Additionally, the performance tradeoff between accuracy and complexity will be examined in terms of computational cost to determine both the advantages and disadvantages of each approach.

In dynamic environments (e.g. an Air Operations Center (AOC)), effective real-time monitoring of mission execution is highly dependent on situation awareness (SA). But whereas an individual's perception of mission progress is biased by his or her immediate tasks and environment, the combined perspectives of key individuals provides an effects-based assessment of the mission overall. Belief networks (BNs) are an ideal tool for modeling and meeting the requirements of SA: at the individual level BNs emulate a skilled human's information fusion and reasoning process in a multi-task environment in the presence of uncertainty. At the mission level, BNs are intelligently combined to yield a common operating picture. While belief networks offer significant advantages for SA in this manner, the work of defining and combining the models is difficult due to factors such as multiple-counting and conflicting reports. To address these issues, we develop a system consisting of three distinct functional elements: an off-line mechanism for rapid construction of a BN library of SA models tailored to different air combat operation situations and derived from knowledge elicitation with subject matter experts; an off-line mechanism to adapt and combine BN models that supports the ability to adjust the SA models over time and in response to novel situations not initially available or anticipated during model construction; and an on-line combination of SA models to support an enhanced SA and the ability to monitor execution status in real time and informed by and responsive to the individuals and situations involved.

In this paper, the combination rules, such as the Dempster-Shafer's (D-S) combination rule, the Yager's combination rule, the Dubois and Prade's (D-P) combination rule, the DSm's combination rule and the disjunctive combination rule, are applied to the situation assessment and target identification problems. Given two independent sources of information with different resolutions, the results from each combination rule of evidence are analyzed. It is observed from these results that the DSm's rule is the fastest in arriving at a decision compared to the other three rules, while the disjunctive combination rule is the slowest. The Yager's rule yields the same identification results for the situation assessment as the Dubois and Prade's rule. Moreover, the decision-making of the D-S' rule is faster than that of the Yager's as well as of the Dubois and Prade's rules, however, slower than that of the DSm's rule

The importance of network science to the present and future military is unquestioned. Networks of some type pervade every aspect of military operations-a situation that is shared by civilian society. However, several aspects of militarily oriented network science must be considered unique or given significantly greater emphasis than their civilian counterparts. Military, especially battlespace, networks must be mobile and robust. They must utilize diverse sensors moving in and out of the network. They must be able to survive various modes of attack and the destruction of large segments of their structure. Nodes often must pass on classifications made locally while other nodes must serve as combined sensor/classifiers or information coordinators. They must be capable of forming fluidly and in an ad hoc manner. In this paper, it will be shown how category theory, higher category theory, and topos theory provide just the model required by military network science. Category theory is a well-developed mathematical field that views mathematical structures abstractly, often revealing previously unnoticed correspondences. It has been used in database and software modeling, and in sensor and data fusion. It provides an advantage over other modeling formalisms both in its generality and in its extensive theory. Higher category theory extends the insights of category theory into higher dimensions, enhancing robustness. Topos theory was developed, in part, through the application of category theory to logic, but it also has geometric aspects. The motivation behind including topos theory in network science is the idea that a mathematical theory fundamental to geometry and logic should be applicable to the study of systems of spatially distributed information and analysis flow. The structures presented in this paper will have profound and far-reaching applications to military networks.

Key to effective image exploitation is letting man and machine do what they each do best. Automated target recognition (ATR) systems rely on model-based approaches that operate in a highly structured predict-extract-match-search (PEMS) loop. Fundamental to wide area search (WAS) is the problem of detecting a large number of potential objects that are so diverse in nature that they cannot be enumerated let alone modeled. In addition there is the problem of unexpected objects (i.e., those that cannot be modeled a priori). A new approach to search based on a detect-organize-compile (DOC) paradigm is proposed and applied to wide area change detection (WACD). It combines an automated image screening algorithm for detecting manmade changes with an interactive web-based tool for ranking, reviewing, and compiling changes of interest. Performance of this approach against exhaustive manual search of the image shows about a 10X increase in throughput (changes detected per unit time) at a fixed level of performance.

An information-theoretic method is described for automatically determining the best number of clusters. It is motivated by Rissanen's minimum description length principle that states the best representation is the one with the fewest bits. The method is evaluated using two different clustering algorithms: a mode finder based on scale-space algorithm, and a vector quantizer (VQ). Synthetic, single- and multi-band image clustering examples are presented. Clusterings produced by the mode finder are shown to better correspond to distinguishable surface categories in the scene than those produced by the VQ algorithm. VQ clusterings are evaluated within an anomaly detector, which detects manmade object/changes as spectral outliers within a set of background clusters. It is shown that the optimal VQ clustering (the one with the fewest bits) produces the best detection performance.

Current surveillance and reconnaissance systems require improved capability to enable the co-registration of larger images, combining enhanced temporal, spatial, and spectral resolutions. However, such proficient remote sensing systems cannot employ traditional manual exploitation techniques to cope successfully with the avalanche of data to be processed and analyzed. Automated image exploitation tools may be employed if the images are already co-registered together. Therefore, there is a need to develop fully automated co-registration algorithms able to deal with different scenarios, and helpful to be used successively for numerous applications such as image data fusion, change detection, and target detection. This paper describes the Automated Multi-sensor Image Registration (AMIR) system and embedded algorithms under development at DRDC-Valcartier. The AMIR system provides a framework for the automated multi-date registration of electro-optic images, acquired from different sensors and from dissimilar oblique view angles. The system is characterized by its fully automated nature, where no user intervention prevailed. Advanced image algorithms are used in order to supply the capability to register multi-date electro-optic images acquired from different viewpoints, under singular operational conditions, multiple scenarios (e.g. airport, harbor, vegetation, urban, etc.), different spatial resolutions (e.g. IKONOS/QuickBird, Airborne/Spaceborne), while providing sub-pixel accuracy registration level.

Accurate image registration is critical for applications such as precision targeting, geo-location, change-detection, surveillance, and remote sensing. However, the increasing volume of image data is exceeding the current capacity of human analysts to perform manual registration. This image data glut necessitates the development of automated approaches to image registration, including algorithm parameter value selection. Proper parameter value selection is crucial to the success of registration techniques. The appropriate algorithm parameters can be highly scene and sensor dependent. Therefore, robust algorithm parameter value selection approaches are a critical component of an end-to-end image registration algorithm. In previous work, we developed a general framework for multisensor image registration which includes feature-based registration approaches. In this work we examine the problem of automated parameter selection. We apply the automated parameter selection approach of Yitzhaky and Peli to select parameters for feature-based registration of multisensor image data. The approach consists of generating multiple feature-detected images by sweeping over parameter combinations and using these images to generate estimated ground truth. The feature-detected images are compared to the estimated ground truth images to generate ROC points associated with each parameter combination. We develop a strategy for selecting the optimal parameter set by choosing the parameter combination corresponding to the optimal ROC point. We present numerical results showing the effectiveness of the approach using registration of collected SAR data to reference EO data.

Sufficient conditions for strictly positive definite correlation functions are developed. These functions are associated with wide-sense stationary stochastic processes and provide practical models for various errors affecting tracking, fusion, and general estimation problems. In particular, the expected magnitude and temporal correlation of a stochastic error process are modeled such that the covariance matrix corresponding to a set of errors sampled (measured) at different times is positive definite (invertible) - a necessary condition for many applications. The covariance matrix is generated using the strictly positive definite correlation function and the sample times. As a related benefit, a large covariance matrix can be naturally compressed for storage and dissemination by a few parameters that define the specific correlation function and the sample times. Results are extended to wide-sense homogeneous multi-variate (vector-valued) random fields. Corresponding strictly positive definite correlation functions can statistically model fiducial (control point) errors including their inter-fiducial spatial correlations. If an estimator does not model correlations, its estimates are not optimal, its corresponding accuracy estimates (a posteriori error covariance) are unreliable, and it may diverge. Finally, results are extended to approximate error covariance matrices corresponding to non-homogeneous, multi-variate random fields (a generalization of non-stationary stochastic processes). Examples of strictly positive definite correlation functions and corresponding error covariance matrices are provided throughout the paper.

Respiratory waveforms and their derived respiratory rate time-series data can become misaligned from each other when they are collected by vital signs monitors under sub-optimal field conditions. The monitor-provided waveforms and rates can be re-aligned by independently calculating respiratory rates from the waveforms and then aligning them with the monitor-provided rates. However, substantially different rates may be generated from the same waveform due to the presence of ambiguous breaths at noisy positions in the waveform. This paper reports a landscape matching (LAM) algorithm to align respiratory rate time-series data with the waveform that they are derived from by using rates that are calculated by different means. The algorithm exploits the intermittent matches between two respiratory rate time series to generate a matching score for an alignment. The best alignment exhibits the highest matching score. The alignment performance of the LAM algorithm is compared to that of a correlation matching (CM) algorithm using field-collected respiratory data. Alignment performance is evaluated by: (1) comparing the ability of the two algorithms to return a shifted waveform to its original, known position; and (2) comparing the percent of points that match between the monitor-provided and calculated respiratory rate time-series data after re-alignment. The LAM alignment algorithm outperforms the CM algorithm in both comparisons at a statistically significant level (p<0.05). Out of 67 samples with shifted time-series data, on average, the LAM aligns respiratory rates within 44 seconds of the original position, which is significantly better the CM-calculated alignment (136 seconds). Out of 465 samples, the LAM performs better, worse, and equal to the CM algorithm in percentage of points matching in 73%, 11%, and 16% of the cases, respectively. This robust alignment algorithm supports the use of reliable post-hoc monitor-provided respiratory rates for data mining purposes.

Reliable and accurate methods for detection and extraction of linear network, such as road networks, in satellite imagery are essential to many applications. We present an approach to the road network extraction from high-resolution satellite imagery that is based on proximity graph analysis. We are jumping off from the classification provided by existing spectral and textural classification tools, which produce a set of candidate road patches. Then, constrained Delaunay triangulation and Chordal Axis transform are used to extract centerline characterization of the delineated candidate road patches. We refine produced center lines to reduce noise influence on patch boundaries, resulting in a smaller set of robust center lines authentically representing their road patches. Refined center lines are triangulated using constrained Delaunay triangulation (CDT) algorithm to generate a sub-optimal mesh of interconnections among them. The generated triangle edges connecting different center lines are used for spatial analysis of the center lines relations. A subset of the Delaunay tessellation grid contains the Euclidian Minimum Spanning Tree (EMST) that provides an approximation of road network. The approach can be generalized to the multi-criteria MST and multi-criteria shortest path algorithms to integrate other factors important for road network extraction, in addition to proximity relations considered by standard EMST.

We developed a method to fuse imagery for from different sensor types. The core of our method uses different forward transforms of input images and a common transform to reconstruct the final result. When measuring the entropy and power of the fused result, we found that our method gave generally better results when compared to a more conventional approach. Our method could form the basis of a new image fusion approach because it offers results not possible with a conventional approach.

Ground moving target tracking in aerial video presents a difficult algorithmic challenge due to sensor platform motion, non-uniform scene illumination, and other extended operating conditions. Theoretically, trackers which operate on color video should have improved performance vs. monochromatic trackers by leveraging the additional intensity channels. In this work, ground moving targets in color video are characterized in the Hue-Saturation-Value (HSV) color space. Using segmented real aerial video, HSV statistics are measured for multiple vehicle and background types and evaluated for separability and invariance to illumination change, obscuration, and aspect change. HSV statistics are then calculated for moving targets from the same video segmented with existing color tracking algorithms to determine HSV feature robustness to noisy segmentation.

In recent years, active research has mainly concentrated on authenticating a signature; tracking a document in a digital library, and tamper detection of a scanned document or secured communication using binary images. Binary image steganographical systems provide a solution for the above discussed issues. The two color constraint of the image limits the extension of various LSB embedding techniques to the binary case. In this paper, we present a new data hiding system for binary images and scanned documents. The system initially identifies embeddable blocks and enforces specific block statistics to hide sensitive information. The distribution of the flippable pixels in these blocks is highly uneven over the image. A variable block embedding threshold is employed for capitalizing on this uneven distribution of pixels. In addition, we also present a measure to find the best the cover given a specific file of sensitive information. The simulation was performed over 50 various binary images such the scanned documents, cartoons, threshold color images. Simulation results shows that 1) The amount of data embedded is comparatively higher than the existing algorithms (such as K.H. Hwang et.al [5], J. Chen et.al [10], M.Y.Wu et.al [9]). 2) The visual distortion in cover image is minimal when compared with the existing algorithms (such as J. Chen[10], M.Y.Wu et.al [9]) will be presented.

Modern computers can be made more friendly and otherwise improved by making them behave more like humans. Perhaps we can learn how to do this from biology in which human brains evolved over a long period of time. Therefore, we first explain a commonly used biological neural network (BNN) model, the Wilson-Cowan neural oscillator, that has cross-coupled excitatory (positive) and inhibitory (negative) neurons. The two types of neurons are used for frequency modulation communication between neurons which provides immunity to electromagnetic interference. We then evolve, for the first time, an artificial neural network (ANN) to perform the same task. Two dynamical feed-forward artificial neural networks use cross-coupling feedback (like that in a flip-flop) to form an ANN nonlinear dynamic neural oscillator with the same equations as the Wilson-Cowan neural oscillator. Finally we show, through simulation, that the equations perform the basic neural threshold function, switching between stable zero output and a stable oscillation, that is a stable limit cycle. Optical implementation with an injected laser diode and future research are discussed.

A significant data quality challenge for highly variant systems surrounds the limited ability to quantify operationally reasonable limits on the data elements being collected and provide reasonable threshold predictions. In many instances, the number of influences that drive a resulting value or operational range is too large to enable physical sampling for each influencer, or is too complicated to accurately model in an explicit simulation. An alternative method to determine reasonable observation thresholds is to employ an automation algorithm that would emulate a human analyst visually inspecting data for limits. Using the visualization technique of self-organizing maps (SOM) on data having poorly understood relationships, a methodology for determining threshold limits was developed. To illustrate this approach, analysis of environmental influences that drive the abundance of a target indicator species (the pink shrimp, Farfantepenaeus duorarum) provided a real example of applicability. The relationship between salinity and temperature and abundance of F. duorarum is well documented, but the effect of changes in water quality upstream on pink shrimp abundance is not well understood. The highly variant nature surrounding catch of a specific number of organisms in the wild, and the data available from up-stream hydrology measures for salinity and temperature, made this an ideal candidate for the approach to provide a determination about the influence of changes in hydrology on populations of organisms.

This paper will compare the various methods of analyzing the results of radar pulse train deinterleavers. This paper is divided into three sections. The first section of this paper will describe the basic methods, such as the confusion matrix, and some measures that can be obtained from the matrix. The measures will include correct correlation, miscorrelation, ambiguity and track purity. Correct correlation is calculated by dividing the total number of correctly clustered pulses by the total number of pulses in the collect. Miscorrelation measures the fraction of received pulses that incorrectly deinterleaved. Ambiguity measures the fraction of received pulses that are rejected by the deinterleaver as having uncertain association with a ground truth track. Track purity measures the ability of the deinterleaver to create a constructed track comprised of pulses from a single ground truth track. These metrics will show the quality of the deinterleaving operation. The second section of this paper will describe some of the advanced similarity measures of effectiveness. This section will also describe how distance measures will be used to analyze deinterleaver results. The two main similarity measures to be discussed in this paper will be the Rand Adjust and Jaccard coefficient. These similarity measures are also known as criterion indices and are used for evaluating the capacity to recover true cluster structure. The reason for the selection of the Jaccard and Rand Adjust as measures is that they both allow a value to be obtained that is between 0 and 1 that will show how good the clusterer in question has performed. The Rand Adjust also allows for more variability in the range between 0 and 1 and appears to provide a more accurate evaluation of the cluster. The distance measures that will be described include Euclidean, Mahalanobis and Minkowski distances. These distance measures have different methods to evaluate each cluster for purity. These measures will provide an indication of the quality of the deinterleaver operation. The third section of this paper will provide the results from these measures on a deinterleavered dataset and will discuss the comparison of these metrics.

Purity of deinterleaved pulse descriptor word (PDW) trains is critical to the performance of emitter classification software that analyzes PDW data. Contamination of the input PDW train can lead to artifacts in the analysis resulting in incorrect or ambiguous classification. This paper presents results of an investigation into the possibility of applying transition matrices to the detection and removal of contaminating pulses from a deinterleaver output. The utility of transition matrices as a fast and efficient pattern recognition tool in ESM emitter classification has been known for over a decade [Brown, R.G., "Pattern Recognition Using Transition Matrices", 7^th RMC Symposium on Applications of Advanced Software in Engineering, pp 52-57, Royal Military College of Canada, Kingston Ontario, May 1995]. In the work presented herein, transition matrix patterns unique to contaminated pulse trains are sought in order to provide a warning that a particular PDW train is contaminated, and provide a clue as to which pulses should be removed to purify the PDW train.

This paper provides a comparison of the two main techniques currently in use to solve the problem of radar pulse train deinterleaving. Pulse train deinterleaving separates radar pulse trains into the tracks or bins associated with the detected emitters. The two techniques are simple time of arrival (TOA) histogramming and multi-parametric analysis. TOA analysis uses only the time of arrival (TOA) parameter of each pulse to deinterleave radar pulse trains. Such algorithms include Cumulative difference (CDIF) histogramming and Sequential difference (SDIF) histogramming. Multiparametric analysis utilizes any combination of the following parameters: TOA, radio frequency (RF), pulse width (PW), and angle of arrival (AOA). These techniques use a variety of algorithms, such as Fuzzy Adaptive Resonance Theory (Fuzzy-ART), Fuzzy Min-Max Clustering (FMMC), Integrated Adaptive Fuzzy Clustering (IAFC) and Fuzzy Adaptive Resonance Theory Map (Fuzzy-ARTMAP) to compare the pulses to determine if they are from the same emitter. Good deinterleaving is critical since inaccurate deinterleaving can lead to misidentification of emitters. The deinterleaving techniques evaluated in this paper are a sizeable and representative sample of both US and international efforts developed in the UK, Canada, Australia and Yugoslavia. Mardia [1989] and Milojevic and Popovich [1992] shows some of the early work in TOA-based deinterleaving. Ray [1997] demonstrates some of the more recent work in this area. Multi-parametric techniques are exemplified by Granger, et al [1998] and Thompson and Sciortino [2004]. This paper will provide an analysis of the algorithms and discuss the results obtained from the referenced articles. The algorithms will be evaluated for usefulness in deinterleaving pulse trains from agile radars.

This work presents a new methodology for the formulation of discrete chirp Fourier transform (DCFT) algorithms and it discusses performance measures pertaining to the mapping of these algorithms to hardware computational structures (HCS) as well as the extraction of chirp rate estimation parameters of multicomponent nonstationary signals arriving from point targets. The methodology centers on the use of Kronecker products algebra, a branch of finite dimensional multilinear algebra, as a language to present a canonical formulation of the DCFT algorithm and its associated properties. The methodology also explains how to search for variants of this canonical formulation that contribute to enhance the mapping process to a target HCS. The parameter extraction technique uses time-frequency properties of the DCFT in a modeled delay-Doppler synthetic aperture radar (SAR) remote sensing and surveillance environment to treat multicomponent return signals of prime length, with additive Gaussian noise as background clutter, and extract associated chirp rate parameters. The fusion of time-frequency information, acquired from transformed chirp or linear frequency modulated (FM) signals using the DCFT, with information obtained when the signals are treated using the discrete ambiguity function acting as point target response, point spread function, or impulse response, is used to further enhance the estimation process. For the case of very long signals, parallel algorithm implementations have been obtained on cluster computers. A theoretical computer performance analysis was conducted on the cluster implementation based on a methodology that applies well-defined design of experiments methods to the identification of relations among different levels in the process of mapping computational operations to high-performance computing systems. The use of statistics for identification of relationships among factors has formalized the search for solutions to the mapping problem and this approach allows unbiased conclusions about results.

Barker codes are the only known bi-phase codes with the smallest achievable sidelobes. However, the longest Barker code of odd length is proven to be that of length 13. There is also a strong conjecture that the Barker codes of length 2 and 4 are the only ones there are of even length. The code of length 13 achieves a mainlobe to peak sidelobe ratio of only 22.28 dB which is less than the practical requirement of at least 30 dB. Mismatched filters have been proposed in the literature to improve this ratio. This is achieved at the cost of some deterioration in the signal to noise ratio at the filter output. In this paper, very efficient sidelobe suppression of Barker codes of length 13 is achieved through the use of a novel mismatched filter. This mismatched filter is comprised of a conventional matched filter cascaded with a computationally efficient filter based on multiplicative expansion. Several parameters are introduced in the terms of the expansion and are optimized to improve the performance of the filter. A version of this filter with three stages is shown to achieve a mainlobe to peak sidelobe ratio of almost 61.9 dB. This is achieved at the cost of some deterioration in the signal to noise ratio at the filter output. The loss in signal to noise ratio (SNR) is shown to be only 0.2 dB. It is suggested that by applying the proposed technique to compound Barker codes, the mainlobe to peak sidelobe ratio could be maintained while improving the signal to noise ratio performance.

Tropospheric aerosols play an important role in climate change. Aerosols are typically studied over deep clear water, due to the relatively constant reflectance of water and the ability to easily separate surface and atmospheric contributions on the satellite signal. A methodology based on multi-spectral approach was employed to map tropospheric aerosols concentrations over the water areas surrounding Penang Island. The aim of this study was to estimate the pollutants concentrations using remote sensing techniques. In this study, we attempted to derive AOT (Aerosol Optical Thickness) values from the sky transmittance measurements in the visible spectrum. The transmittance values were measured at the sea surface using a handheld spectroradiometer. The correspond PM10 readings were taken simultaneously during the transmittance measurements acquisition of the imageries using a Dust Trak meter. The PCI Geomatica version 9.1 digital image processing software was used in all image-processing analyses. The results produced a linear relationship between PM10 and AOT values over the water surface of Penang Island. Finally, The PM10 concentration map over the water surface of Penang Island was generated using Kriging interpolation technique. This study has indicated the potential use of a handheld spectroradiometer for air quality study.

Invited Panel Discussion Topics: Research Challenges in Sensor Management; Fundamental Statistics for Resource Management; Issues in Formulating Utility Functions for Sensor Management; Resource Management for Distributed Attention in Sensor Networks; Research Challenges in Network and Service Management for Distributed Net-Centric Fusion; Resource Management in Sensor Networks; Performance Metrics for Combed Tracking and Sensor Management.