Taek Song
Hanyang University, Electronic Systems, Faculty Member
A Gaussian Mixture (GM) target tracking solution is a natural consequence of the multi-target tracking in clutter, given a linear target trajectory propagation and a linear target measurement equation. We examine and compare two prominent... more
A Gaussian Mixture (GM) target tracking solution is a natural consequence of the multi-target tracking in clutter, given a linear target trajectory propagation and a linear target measurement equation. We examine and compare two prominent GM target trackers: the Multi Hypothesis Tracking (MHT) and the Integrated Track Splitting. Both incorporate the false track discrimination capabilities, enabling automatic target tracking in the presence of clutter measurements and missed detections.
Research Interests:
Multistatic tracking involves using non-collocated transmitters and receivers to track the targets. In this paper we also assume no angle information. This paper proposes an approach using Gaussian Mixture representation of measurement... more
Multistatic tracking involves using non-collocated transmitters and receivers to track the targets. In this paper we also assume no angle information. This paper proposes an approach using Gaussian Mixture representation of measurement uncertainty. An arbitrary number of receivers can be used using this approach. Here we limit ourselves to the track estimation issues, i.e. we assume no clutter measurement, and imperfect target detection at the receivers. Simulation results are presented which vindicate this approach.
Research Interests:
Research Interests:
The target tracking using multistatic passive radar in a digital audio/video broadcast (DAB/DVB) network with uncertain illuminators of opportunity faces two main challenges: one is the three-dimensional (3-D) data association problem... more
The target tracking using multistatic passive radar in a digital audio/video broadcast (DAB/DVB) network with uncertain illuminators of opportunity faces two main challenges: one is the three-dimensional (3-D) data association problem among the target-measurement-illuminator due to the uncertainty of illuminators of opportunity; the other is that only the bistatic range and range-rate measurements are available since the angle information is unavailable or of very poor quality. In this paper, the authors propose two tracking algorithms directly in a two dimensional (2-D) Cartesian coordinates with the capability of false track discrimination (FTD) using the probability of target existence: the modified joint integrated data association (MJIPDA) and the sequential processing joint integrated data association (SP-JIPDA). The MJIPDA algorithm is enhanced based on the existing MJPDA algorithm by using the probability of target existence for FTD. The SP-JIPDA algorithm sequentially operates the JIPDA tracker to update each track for each illuminator with all the measurements in the common measurement set at each time. A simulation study is performed to verify the validity of both algorithms in this multistatic passive radar system.
Research Interests:
Research Interests: Geology, Computer Science, Algorithm, Monte Carlo Simulations, Monte Carlo Methods, and 14 moreGeolocation, Sensors, Noise, Cramer Rao Lower Bound, Spatialization, Vectors, CRLB, Covariance Matrix, Optimal Bayesian Estimation, Time measurement, Time of Arrival, Multilateration, common emitter, and hyperbola
ABSTRACT Data association attempts to discriminate between the target and the clutter measurements, usually calculating the posterior probabilities of measurement origins. The clutter (spurious) measurements are random and (we presume)... more
ABSTRACT Data association attempts to discriminate between the target and the clutter measurements, usually calculating the posterior probabilities of measurement origins. The clutter (spurious) measurements are random and (we presume) follow the Poisson distribution. The Poisson distribution is non-homogeneous and is parametrized by intensity (the clutter measurement density). The clutter measurement density is almost always a priori unknown, and is often non stationary. Here we propose a measurement oriented clutter density estimator with probability hypothesis density (PHD) filtering which integrates information from single-scan spatial clutter density estimator, and can follow and smooth non-stationary clutter information.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
The point target assumption which allows each target to generate at most one measurement at each scan is widely used in target tracking field. However, in come tracking scenarios, one target can generate many measurements due to high... more
The point target assumption which allows each target to generate at most one measurement at each scan is widely used in target tracking field. However, in come tracking scenarios, one target can generate many measurements due to high sensor resolution which give rise to the multiple detection problem. Traditional algorithms get poor performances since each data association event considers only one measurement as target detection to estimate target state. The measurement partition method which forms all possible combinations of target originated measurements is designed for multiple detection problem. In this paper, joint integrated track splitting (JITS) tracker and measurement partition method are combined to generate a new structure, called multiple detection joint integrated track splitting (MD-JITS) tracker, to extract and utilize the target motion information contained in the measurements more effectively. Compared with some existing filters, this filter achieves better tracking performance for automatic false track discrimination (FTD).
Research Interests:
Research Interests:
A radar system for tracking ground vehicle targets to realize adaptive cruise control requires an accurate vehicle tracking filter. Especially, for ground vehicle tracking, one has to consider the case in which the radar measurements are... more
A radar system for tracking ground vehicle targets to realize adaptive cruise control requires an accurate vehicle tracking filter. Especially, for ground vehicle tracking, one has to consider the case in which the radar measurements are affected by glint noises generated by the targets located near the observer vehicle. Furthermore, this vehicle tracking should be performed in cluttered environments. This paper presents a combined algorithm that consists of integrated probabilistic data association (IPDA), and an interacting multiple model (IMM) algorithm for ground target tracking in clutter and target glint. Performance of the proposed algorithm is tested and verified by a series of computer simulation runs.
Research Interests:
Research Interests:
Research Interests:
ABSTRACT We consider passive surveillance using time and frequency difference of arrival signals received by mobile receiver pairs. Signals received by a pair of receivers are correlated in time and frequency, followed by a detection... more
ABSTRACT We consider passive surveillance using time and frequency difference of arrival signals received by mobile receiver pairs. Signals received by a pair of receivers are correlated in time and frequency, followed by a detection process. In addition to the target (emitter) measurements, we may also create a number of spurious detections in each scan. This paper considers local (distributed) tracking using these measurements, with the main purpose of eliminating spurious measurements and enhancing the emitter detection.
Research Interests:
This paper presents an algorithm for multistatic target tracking in clutter, using only range difference information (neither bearing nor Doppler information are assumed available). Presence of false tracks, Data association issues as... more
This paper presents an algorithm for multistatic target tracking in clutter, using only range difference information (neither bearing nor Doppler information are assumed available). Presence of false tracks, Data association issues as well as the nonlinear measurement equation makes this a challenging problem. This paper proposes a solution to this problem by using the Gaussian Mixture Measurement likelihood — Integrated Track Splitting algorithm.
Research Interests:
Research Interests:
The authors extend the smoothing integrated probabilistic data association algorithm to multi-target tracking in clutter, or alternatively, use smoothing to improve the joint integrated probabilistic data association (JIPDA) algorithm.... more
The authors extend the smoothing integrated probabilistic data association algorithm to multi-target tracking in clutter, or alternatively, use smoothing to improve the joint integrated probabilistic data association (JIPDA) algorithm. The predictions of forward and backward JIPDA are fused to form the smoothing prediction, which is used for smoothing multi-target data association.
Research Interests:
Research Interests:
Research Interests:
This paper presents an algorithm for high pulserepetition frequency (HPRF) radar tracking in clutter. A most serious problem with HPRF radar is the range ambiguity. This paper proposes a solution to this problem by using the Gaussian... more
This paper presents an algorithm for high pulserepetition frequency (HPRF) radar tracking in clutter. A most serious problem with HPRF radar is the range ambiguity. This paper proposes a solution to this problem by using the Gaussian Mixture Model - Integrated Track Splitting (GMM-ITS) algorithm, which also gives out the probability of target existence (PTE) as the track quality measure. This track quality measure is used for false track discrimination in clutter.
Research Interests:
Research Interests:
Research Interests:
Abstract- The measurement with the strongest signal amplitude in the validation gate is known as the strongest neighbor (SN) measurement. A standard Kalman filter that utilizes the SN at any time as if it is originated from the true... more
Abstract- The measurement with the strongest signal amplitude in the validation gate is known as the strongest neighbor (SN) measurement. A standard Kalman filter that utilizes the SN at any time as if it is originated from the true target is called the strongest neighbor filter(SNF). Inconsistency of handling the SN as if it is true target is corrected in the existing probabilistic strongest neighbor filter(PSNF) which accounts the probability that the SN is from the true target. It is known that performance of the PSNF is superior to the SNF at a cost of increased computational load. In this paper, we propose a new probabilistic strongest neighbor filter that takes into account the current number of validated measurements in the derivation of probability density functions for the SN which are needed to establish probability weightings and estimation error covariance. The proposed algorithm does not involve infinite summation while the existing PSNF algorithm contains infinite summ...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
The Integrated Probabilistic Data Association (IPDA) filter is derived using an analytic combinatorial method. The IPDA filter and the underlying discrete-continuous event space are unchanged. The feasible measurement assignments are... more
The Integrated Probabilistic Data Association (IPDA) filter is derived using an analytic combinatorial method. The IPDA filter and the underlying discrete-continuous event space are unchanged. The feasible measurement assignments are encoded by a probability generating functional (PGFL). The probability distributions are decoded by the derivatives of the PGFL. The IPDA derivation is analytic in the sense that the enumeration of feasible assignments is implicit in the derivatives of the PGFL. The Joint IPDA (JIPDA) filter assumes-like its classical JPDA counterpart-that each target has its own state space. This important modeling feature distinguishes the JIPDA from a related family of filters based on multi-Bernoulli point processes that are superimposed in one targets state space.
Research Interests: Mathematics and IEEE
This article presents a method for depth estimation using image focus based on the linear regression model. Two datasets are selected for each pixel based on the maximum value which is calculated using Laplacian operator. Then linear... more
This article presents a method for depth estimation using image focus based on the linear regression model. Two datasets are selected for each pixel based on the maximum value which is calculated using Laplacian operator. Then linear regression model is used to find lines that approximate these datasets. The best fit lines are found using least squares method. After approximating the two lines, their intersection point is calculated, and weights are assigned to calculate the new value for the depth map. The proposed method is compared with four depth estimation algorithms. Six different objects are selected for testing the proposed method. © 2011 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 21, 241–246, 2011;
Research Interests:
Research Interests:
Multi-target tracking (MTT) is a challenging issue due to an unknown number of real targets, motion uncertainties, and coalescence behavior of sensor (such as radar) measurements. The conventional MTT systems deal with intractable... more
Multi-target tracking (MTT) is a challenging issue due to an unknown number of real targets, motion uncertainties, and coalescence behavior of sensor (such as radar) measurements. The conventional MTT systems deal with intractable computational complexities because they enumerate all feasible joint measurement-to-track association hypotheses and recursively calculate the a posteriori probabilities of each of these joint hypotheses. Therefore, the state-of-art MTT system demands bypassing the entire joint data association procedure. This research work utilizes linear multi-target (LM) tracking to treat feasible target detections followed by neighbored tracks as clutters. The LM integrated track splitting (LMITS) algorithm was developed without a smoothing application that produces substantial estimation errors. Smoothing refines the state estimation in order to reduce estimation errors for an efficient MTT. Therefore, we propose a novel Fixed Interval Smoothing LMITS (FIsLMITS) algor...
Research Interests:
Multi static underwater surveillance must conserve energy by communicating between nodes as sparsely as possible. On the other hand, signal to clutter ratio is usually very low, resulting in large number of clutter measurements per ping.... more
Multi static underwater surveillance must conserve energy by communicating between nodes as sparsely as possible. On the other hand, signal to clutter ratio is usually very low, resulting in large number of clutter measurements per ping. When the local target tracking is performed in each receiver node, the false measurements are largely eliminated and communications are used (almost) exclusively to report targets to the fusion center.
Research Interests:
Research Interests:
Abstract: The template matching method is used as a simple method to track objects or patterns that we want to search for in the input image data from image sensors. It recognizes a segment with the highest correlation as a target. The... more
Abstract: The template matching method is used as a simple method to track objects or patterns that we want to search for in the input image data from image sensors. It recognizes a segment with the highest correlation as a target. The concept of this method is similar to that of SNF (Strongest Neighbor Filter) that regards the measurement with the highest signal intensity as target-originated among other measurements. The SNF assumes that the strongest neighbor (SN) measurement in the validation gate originates from the target of interest and the SNF utilizes the SN in the update step of a standard Kalman filter (SKF). The SNF is widely used along with the nearest neighbor filter (NNF), due to computational simplicity in spite of its inconsistency of handling the SN as if it is the true target. Probabilistic Strongest Neighbor Filter for m validated measurements (PSNF-m) accounts for the probability that the SN in the validation gate originates from the target while the SNF assumes...
Research Interests:
Clutter measurement density (CMD) is one of data association parameters, which indicates the number of clutter measurements per unit volume of the measurement space. In probabilistic data association based algorithms, the association... more
Clutter measurement density (CMD) is one of data association parameters, which indicates the number of clutter measurements per unit volume of the measurement space. In probabilistic data association based algorithms, the association probability between a prior estimate and a measurement is proportional to the ratio of target measurement likelihood and CMD. Also the measurement likelihood is used for obtaining the target existence probability for false track discrimination. Although CMD is an important parameter for state estimation as well as track management, it depends on surveillance environments in which the true CMD is rarely known in advance. A clutter measurement density estimator (CMDE) calculates the spatial density of clutter adaptively using measurement information, and provides its estimated CMD to data association algorithms for adaptive target tracking in clutter. A spatial CMDE (SCMDE) selects the measurement with the N-th smallest 2-norm distance from the measuremen...
Research Interests:
Target tracking in surveillance, guidance, radar, sonar systems plays an important role for prediction and estimation of target's information. In reality, target tracking is usually attended in imperfect target detection environment... more
Target tracking in surveillance, guidance, radar, sonar systems plays an important role for prediction and estimation of target's information. In reality, target tracking is usually attended in imperfect target detection environment and in cluttered environment caused by false alarm, multipath fading, and non-specific noise. Target tracking algorithms usually treat the probability of detection as independent of the target state. In most cases, this assumption is not true, with subsequent degradation in the target tracking performance from both expected and optimal levels. One typical example is the Doppler frequency based clutter rejection, the other is obfuscation (shadowing) of ground based targets, and the third is anti-jamming notch filtering. In this paper, we present an algorithm for multi-target tracking with target state dependent detection.
Research Interests:
Target tracking in clutter is usually evaluated using simulations. The statistics of the overall tracking performance are gathered, including the false track discrimination, the root mean square trajectory tracking errors, or some... more
Target tracking in clutter is usually evaluated using simulations. The statistics of the overall tracking performance are gathered, including the false track discrimination, the root mean square trajectory tracking errors, or some combined measure of performance, e.g. OSPA. These performance measures treat the target tracking filter as a black box; i.e. the performance of separate target tracking algorithm components is neither considered nor evaluated. In this paper we partition the target tracking algorithm into the estimation and the data association components and consider the performance measures for the data association component. Three performance measures are proposed and evaluated: the fidelity measure, the selection measure and the competence measure which combines capabilities of the fidelity and the selection measures.
Research Interests:
A practical probabilistic data association filter is proposed for tracking multiple targets in clutter. The number of joint data association events increases combinatorially with the number of measurements and the number of targets, which... more
A practical probabilistic data association filter is proposed for tracking multiple targets in clutter. The number of joint data association events increases combinatorially with the number of measurements and the number of targets, which may become computationally impractical for even small numbers of closely located targets in real target-tracking applications in heavily cluttered environments. In this paper, a Markov chain model is proposed to generate a set of feasible joint events (FJEs) for multiple target tracking that is used to approximate the multi-target data association probabilities and the probabilities of target existence of joint integrated probabilistic data association (JIPDA). A Markov chain with the transition probabilities obtained from the integrated probabilistic data association (IPDA) for single-target tracking is designed to generate a random sequence composed of the predetermined number of FJEs without incurring additional computational cost. The FJEs gene...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
A target angular information in 3-dimensional space consists of an elevation angle and azimuth angle. Acoustic signals propagating along multiple paths in underwater environments usually have different elevation angles. Target motion... more
A target angular information in 3-dimensional space consists of an elevation angle and azimuth angle. Acoustic signals propagating along multiple paths in underwater environments usually have different elevation angles. Target motion analysis (TMA) uses the underwater acoustic signals received by a passive horizontal line array to track an underwater target. The target angle measured by the horizontal line array is, in fact, a conical angle that indicates the direction of the signal arriving at the line array sonar system. Accordingly, bottom bounce paths produce inaccurate target locations if they are interpreted as azimuth angles in the horizontal plane, as is commonly assumed in existing TMA technologies. Therefore, it is necessary to consider the effect of the conical angle on bearings-only TMA (BO-TMA). In this paper, a target conical angle causing angular ambiguity will be simulated using a ray tracing method in an underwater environment. A BO-TMA method using particle swarm o...
Research Interests:
The point detections obtained from radars or sonars in surveillance environments include clutter measurements, as well as target measurements. Target tracking with these data requires data association, which distinguishes the detections... more
The point detections obtained from radars or sonars in surveillance environments include clutter measurements, as well as target measurements. Target tracking with these data requires data association, which distinguishes the detections from targets and clutter. Various algorithms have been proposed for clutter measurement density estimation to achieve accurate and robust target tracking with the point detections. Among them, the spatial clutter measurement density estimator (SCMDE) computes the sparsity of clutter measurement, which is the reciprocal of the clutter measurement density. The SCMDE considers all adjacent measurements only as clutter, so the estimated clutter measurement density is biased for multi-target tracking applications, which may result in degraded target tracking performance. Through the study in this paper, a major source of tracking performance degradation with the existing SCMDE for multi-target tracking is analyzed, and the use of the clutter measurement p...