Extended Target Tracking Research Articles

Improved generalized labeled multi-Bernoulli filter for non-ellipsoidal extended targets or group targets tracking based on random sub-matrices

For non-ellipsoidal extended targets and group targets tracking, using an ellipsoid to approximate the target extension may not be accurate enough because of lacking shape and orientation information. In consideration of this, we model a non-ellipsoidal extended target or target group as a combination of multiple ellipsoidal sub-objects, each represented by a random matrix. Besides, in practical tracking scenes, the knowledge of detection probability is of critical importance, which, however, is prior unknown due to the fluctuation of target parameters. Thus based on the aforementioned random sub-matrices model, an improved generalized labeled multi-Bernoulli (GLMB) filter is presented for the tracking of non-ellipsoidal extended targets or group targets in the case of unknown detection probability. In addition, a cheaper Gibbs sampling based joint prediction and updating implementation of the proposed GLMB filter is given, within which the extended targets or group targets are modeled using Beta gamma Gaussian inverse Wishart (BGGIW) distributions, achieving the estimation to detection probability, measurement rate, and extension state of each sub-object and the kinematic state with the positions of all sub-objects and unified kinematics for each extended target or target group. The simulation results demonstrate the effectiveness of the proposed filter.

Distributed Asynchronous Extended Target Tracking Using Random Matrix

In multiple sensor extended target tracking problems, asynchronous measurements are inevitable, since sensors usually have distinct sampling rates and initial sampling times. This paper presents a new distributed extended target tracking algorithm with asynchronous measurements for multiple sensor scenarios. A distributed Bayesian estimation scheme for asynchronous measurements using random matrix framework is derived. We also proposed an effective implementation using particle filtering. Compressed Gaussian Mixture approximations of extended state distributions are exchanged and fused between neighbor sensors. The temporal evolution of elliptic extent parameters can be obtained explicitly in our algorithm. Simulations show reasonable performance with a significant reduction of communication costs for small size systems compared with the centralized algorithm.

A Closed-Form Prediction Update for Extended Target Tracking Using Random Matrices

This paper proposes a new class of state transition models that afford closed-form predictions for the tracking of extended targets. A key innovation is to employ a non-central inverse Wishart distribution to model the state transition density of the target extent. Importantly, this results in a simplified prediction update that is computationally efficient and improves target tracking performance when compared to state-of-the-art alternatives on standard simulation scenarios.

Extended target tracking for autonomous street crossing

Abstract Autonomous navigation on sidewalks and pedestrian areas is a complex problem, that requires the solution of different challenging tasks. One that is particularly hard to tackle is that of autonomous street crossing, which requires the robot to be aware of the position and speed of surrounding vehicles in order to decide whether is safe to cross. This work is dedicated to the development of an obstacle speed estimation algorithm to be applied to the context of autonomous navigation at crosswalks. In particular, a novel approach to the extended-target tracking problem is presented, which leverages a nested structure and a clustering algorithm that reduces the problem to a standard target tracking one. The effectiveness of the algorithm is demonstrated through testing on a prototype parcel-delivery robot operating in a real-world urban environment.

Accurate Tracking of Manoeuvring Target using Scale Estimation and Detection

Camera zoom operation and fast approaching/receding target causes scaling of acquired target in video frames. Fast moving target manifests in large inter-frame motion. In general, non-uniform background degrades performance of tracking algorithms. Fast Fourier transform (FFT)-based Correlation algorithms improve tracking in this scenario, but their applications is limited to small inter-frame motion. Increasing search region has implication on execution speed of the algorithms. Rapid target scaling, non-uniform background and large inter-frame motion of target hinder accurate and long term visual tracking. These challenges have been addressed for extended target tracking by augmenting fast discriminative scale space tracking (fDSST) algorithm with probable target location prediction and target detection. Localisation of fast motion has been achieved by applying fused outputs of Kalman filter and quadratic regression based prediction before applying fDSST. It has helped in accurate localisation of fast motion without increasing search region. In each frame, target location and size have been estimated using fDSST and further refined by target detection near this location. Smoothing and limiting of trajectory and size of detected target has enhanced tracking performance. Experimental results show considerable improvement of precision, success rate and centre location error tracking performance against state-of-the-art trackers in stringent conditions.

Seamless Tracking of Apparent Point and Extended Targets Using Gaussian Process PMHT

In practical target tracking scenarios, targets of different sizes (or extents) may be near or far away from the sensor, which may result in targets appearing as point sources or as extended targets spanning one or more resolution cells, respectively, depending on distance and sensor resolution. In this paper, a new Gaussian Process (GP) measurement model is proposed to explicitly describe the observation about each basis point of GP by an individual dynamic Poisson measurement rate. By employing this model, a novel algorithm to track multiple point targets and extended targets, simultaneously and seamlessly, in the presence of clutter and missed detections is proposed within the Probabilistic Multi-Hypothesis Tracker (PMHT) framework. The proposed algorithm can adapt to spatio-temporally varying target sizes or extents of extended targets and temporally varying target cardinality. In addition, the posterior Cramer-Rao lower bound (PCRLB) for extended targets, which quantifies the accuracies of estimates of multiple extended target states in scenarios with clutter, is derived. Simulations with a scenario consisting of multiple extended targets and point targets are used to verify the effectiveness of the proposed algorithm and to compare its performance with the extended target PCRLB and with those of existing extended target tracking algorithms.

GP-PDA Filter for Extended Target Tracking With Measurement Origin Uncertainty

Extended target tracking (ETT) is an issue in high-resolution radar surveillance, ship tracking, and video tracking. Most of the previous works focus on tracking an ellipsoidal extended target without measurement origin uncertainty (missed detections and clutter). In this paper, a new estimator called the Gaussian process probabilistic data association (GP-PDA) filter is proposed to track an irregularly shaped extended target with measurement origin uncertainty. First, a generalized measurement model for ETT using the Gaussian process (GP) is presented. Both the interior scattering points and the external clutter are considered in this model. Second, a GP-based gating technique is constructed to select validated measurements to feed the filter. Third, the GP-PDA filter is proposed to simultaneously estimate the kinematic state and the contour state of the extended target with measurement origin uncertainty. It is proven that the GP-PDA is a generalized version of the classic PDA, and the latter is a special case of the former in the point target tracking applications. Finally, the GP-based posterior Cramer–Rao lower bound (PCRLB) is derived to evaluate the performance of the ETT with measurement origin uncertainty. Two cases of the PCRLB are discussed, with the number of scattering points being known and unknown. Simulation results verify the effectiveness of the proposed method.

Network Flow Labeling for Extended Target Tracking PHD Filters

The probability hypothesis density (PHD) filter is a method for tracking multiple target objects based on unlabeled detections. However, as the PHD filter employs a first-order approximation of random finite sets, it does not provide track labels, i.e., targets of consecutive time steps are not associated with each other. In this paper, an intuitive and efficient labeling strategy on top of the extended target PHD filter is proposed. The approach is based on solving a network flow problem and makes use of the Wasserstein metric to account for the spatial extent of the objects. The resulting tracker is evaluated with laser scanner data from two traffic scenarios.

Fast density peak-based clustering algorithm for multiple extended target tracking

The key challenge of the extended target probability hypothesis density (ET-PHD) filter is to reduce the computational complexity by using a subset to approximate the full set of partitions. In this paper, the influence for the tracking results of different partitions is analyzed, and the form of the most informative partition is obtained. Then, a fast density peak-based clustering (FDPC) partitioning algorithm is applied to the measurement set partitioning. Since only one partition of the measurement set is used, the ET-PHD filter based on FDPC partitioning has lower computational complexity than the other ET-PHD filters. As FDPC partitioning is able to remove the spatially close clutter-generated measurements, the ET-PHD filter based on FDPC partitioning has good tracking performance in the scenario with more clutter-generated measurements. The simulation results show that the proposed algorithm can get the most informative partition and obviously reduce computational burden without losing tracking performance. As the number of clutter-generated measurements increased, the ET-PHD filter based on FDPC partitioning has better tracking performance than other ET-PHD filters. The FDPC algorithm will play an important role in the engineering realization of the multiple extended target tracking filter.

Multiple detection joint integrated track splitting for multiple extended target tracking

The extended target tracking problem, which occurs due to a high-resolution sensor resolving multiple scattering points of a target, is considered in this paper. Typical target tracking algorithms which use the point target assumption yield poor performances in extended target tracking scenarios, especially in the presence of closely spaced targets. This performance degeneration is mainly caused by a number of measurements that are shared among tracks when the extended targets are closely spaced. In order to achieve robust tracking performances in multiple extended target tracking environments, a novel tracker, named multiple detection joint integrated track splitting (MD-JITS), is proposed here. This tracker incorporates the measurement cells and the random matrix into a target existence based multitarget track splitting structure. In MD-JITS, each track builds a type of tree, and each branch of the tree is a track component corresponding to a sequence of data associations for multiple scans. The track kinematic state and extension state are obtained on the basis of the track components. Then, the proposed algorithm is tested using both Monte Carlo simulations and a real video surveillance application to demonstrate the effectiveness.

An improved partitioning algorithm based on FCM algorithm for extended target tracking in PHD filter

Extended target Gaussian mixture probability hypothesis density (ET-GM-PHD) filter is a promising approach for multi-extended target tracking problem. Correct partitions of measurements are significant to the GM-PHD filter. In order to partition the measurement sets accurately and rapidly, an improved Fuzzy C-Means (FCM) approach is proposed. The proposed approach consists of two stages. Firstly, elliptical gating and Gaussian kernel density analysis technique are used to remove the clutter. Secondly, an improved FCM approach which takes the full merits of the prior information of targets is implemented. The predictive state of targets is utilized in the FCM initialization. And the shape of targets is considered in the iteration of FCM. Both real data and simulated data are performed in this work. Experimentation infers that the proposed method outperforms the others on both effectiveness and computation, especially when multiple closely targets are involved.

Multiellipsoidal extended target tracking with known extent using sequential Monte Carlo framework

In this paper, we consider a variant of the extended target tracking (ETT) problem, namely the multiel- lipsoidal ETT problem. In multiellipsoidal ETT, target extent is represented by multiple ellipses, which correspond to the origin of the measurements on the target surface. The problem involves estimating the target’s kinematic state and solving the association problem between the measurements and the ellipses. We cast the problem in a sequential Monte Carlo (SMC) framework and investigate different marginalization strategies to find an efficient particle filter. Under the known extent assumption, we define association variables to find the correct association between the measurements and the ellipses; hence, the posterior involves both discrete and continuous random variables. By expressing the measurement likelihood as a mixture of Gaussians we derive and employ a marginalized particle filter for the independent association variables without sampling the discrete states. We compare the performance of the method with its alternatives and illustrate the gain in nonstandard marginalization.

Spatio-Temporal Gaussian Process Models for Extended and Group Object Tracking With Irregular Shapes

Extended object tracking has become an integral part of many autonomous systems during the last two decades. For the first time, this paper presents a generic spatio-temporal Gaussian process (STGP) for tracking an irregular and non-rigid extended object. The complex shape is represented by key points and their parameters are estimated both in space and time. This is achieved by a factorization of the power spectral density function of the STGP covariance function. A new form of the temporal covariance kernel is derived with the theoretical expression of the filter likelihood function. Solutions to both the filtering and the smoothing problems are presented. A thorough evaluation of the performance in a simulated environment shows that the proposed STGP approach outperforms the state-of-the-art GP extended Kalman filter approach [N. Wahlstrom and E. Ozkan, “Extended target tracking using Gaussian processes, IEEE Transactions on Signal Processing,” vol. 63, no. 16, pp. 4165–4178, Aug. 2015] with up to $\text{90}\%$ improvement in the accuracy in position, $\text{95}\%$ in velocity and $\text{7}\%$ in the shape, while tracking a simulated asymmetric non-rigid object. The tracking performance improvement for a non-rigid irregular real object is up to $\text{43}\%$ in position, $\text{68}\%$ in velocity, $\text{10}\%$ in the recall, and $\text{115}\%$ in the precision measures.

Cubature Information Gaussian Mixture Probability Hypothesis Density Approach for Multi Extended Target Tracking

In multi-extended target tracking, each target may generate more than one observation. The traditional probability hypothesis density (PHD)-based methods are no longer effective in such scenarios. Recently, the Gaussian mixture PHD approach for the extended target tracking (ET-GM-PHD) has been presented to solve such a problem. The tracking performance of this approach has been restricted by the following disadvantages. First, it only focuses on the linear models. When targets are moving with nonlinear models, it may lead to great tracking errors. Second, the birth intensities are commonly assumed as a prior. In practice, these intensities are always unknown. In order to improve the tracking performance of the traditional ET-GM-PHD approach, a novel extended target tracking approach, namely the ET-cubature information GM-PHD (ET-CIGM-PHD) approach, has been proposed in this paper. To be more specific, we, first, utilize the cubature information filter (CIF) and gating methods to predict and update the Gaussian mixture components of the ET-GM-PHD approach. In the merit of high estimating accuracy of the CIF method, the tracking accuracy of the traditional ET-GM-PHD approach can be significantly improved. Due to the gating method, only part of cells can be used to construct the observation set in the update stage. Thus, the computational load of our approach can be significantly reduced. Then, we propose an adaptive initiating method for the birth intensity initiating. In our method, we utilize the estimated target set to select the most possible partition. Then, we remove cells associated with the estimated targets from the selected partition. The left cells are used to initiate the birth intensity. With the help of the above implementations, the birth intensity can be adaptively initiated. Using such a method, our approach can solve the cases that the prior information of birth intensity is rather little. The simulation results demonstrate the effectiveness of our approach.

Data association for extended target tracking by BP

Data association with great efficiency is an important problem in multiple extended target tracking. It is more challenging than that in point target tracking as the measurements of extended targets are more. This study presents a graphical model formulation of data association for extended target tracking, which is solved by belief propagation (BP) to obtain estimations. To construct the graphical model, a simplified scheme for measurements is proposed to improve the tracking efficiency, which is based on minimal spanning tree algorithm. Then, the data association problem according to the simplified measurement set is presented, solved by BP. Finally, experiment results show the proposed algorithm has advantages over the algorithm based on joint probabilistic data association and the previous data association algorithm, introduced by Vivone and Braca, in terms of robustness and efficiency.

Bayesian Multiple Extended Target Tracking Using Labeled Random Finite Sets and Splines

In this paper, we propose a technique for the joint tracking and labeling of multiple extended targets. To achieve multiple extended target tracking using this technique, models for the target measurement rate, kinematic component, and target extension are defined and jointly propagated in time under the generalized labeled multi-Bernoulli filter framework. In particular, we developed a Poisson mixture variational Bayesian model to simultaneously estimate the measurement rate of multiple extended targets and extended target extension was modeled using B-splines. We evaluated our proposed method with various performance metrics. Results demonstrate the effectiveness of our approach.

Robust adaptive filtering for extended target tracking with heavy‐tailed noise in clutter

A robust adaptive filter is proposed by using the variational Bayesian (VB) inference to extended target tracking with heavy-tailed noise in clutter. An explicit distribution is used to describe the non-Gaussian heavy-tailed noise based on Student's t -distribution. The need for arbitrary decisions is then eliminated, and the robust operation is provided which is less sensitive to extreme observation. Moreover, an approximate measurement update using the analytical techniques of VB methods is derived to approximate the posterior states at each time step. To obtain a more accurate result, clutter estimation is also integrated considering the uncertainty of target tracking in a cluttered environment. The performance of the proposed algorithm is demonstrated with simulated data.

Extended Target GMPHD Filter Based on Mean Shift and Graph Structure

In view of excessive measurements partition number, a large computation load of extended target tracking and leakage estimation when the extended targets cross, an extended target tracking algorithm based on GMPHD with mean shift and graph structure is proposed. Firstly, the kernel density estimation is used to eliminate the clutter measurements. Secondly, mean shift algorithm is adopted to divide the extended target measurements set, and sub-division is considered to carry or not based on the information fed back from the updated graph structure. Then, the extended target GMPHD algorithm is used to filter. Finally, the graph structure is updated by the one-step predicted value of the filtering result, and the updated graph structure information is used to guide the measurement partition at the next moment. Matlab simulation shows that the algorithm proposed decreases largely the number of measurements partition, reduces the computational complexity, and solves the leakage estimation problem when the targets cross.

An ellipse extended target CBMeMBer filter using gamma and box-particle implementation

This work presents a gamma box-particle implementation of the ellipse extended target cardinality balanced multi-target multi-Bernoulli (ET-CBMeMBer) filter, named the ellipse extended target gamma box particle CBMeMBer (EET-GBP-CBMeMBer) filter, where the ellipse extended targets are characterized using a Poisson model suggested by Gilholm et al. The proposed filter is capable of dynamically tracking multiple ellipse extended targets and estimating the target kinematics and extents as well as the measurement rates of targets, in the presence of clutter measurements, false alarms and missed detections. To get the CBMeMBer recursion of the EET-GBP-CBMeMBer filter, a suitable measurement likelihood is derived for a given partitioning cell, and the main gamma box-particle implementation are given along with the necessary box manipulations and approximations. The capabilities and limitations of the proposed EET-GBP-CBMeMBer filter are demonstrated by simulation examples. The simulation results show that a gamma box-particle implementation of the ET-CBMeMBer filter can avoid the high number of particles and reduce computational burden, compared to a particle implementation of that for extended target tracking.

Augmented state PHD filter for extended target tracking with bias compensation

The extended target probability hypothesis density (ET-PHD) filter developed recently is an elegant algorithm for tracking an unknown number of extended targets based on the random finite set (RFS) theory. However, the standard ET-PHD filter can not be straightforwardly applied to track multiple extended targets in the scenarios with sensor biases. Otherwise, different levels of performance degradation can be easily produced due to the presence of biases in sensor measurements. To solve such a problem, an augmented state ET-PHD filter, in which the joint intensity function of target state and sensor bias is recursively estimated, is proposed in this paper. Under linear Gaussian assumptions on the system models and with the joint intensity function represented by the Gaussian mixtures, a closed-form solution to the augmented state ET-PHD filter is also presented. The simulated results for various clutter rates and Poisson rates of measurements demonstrate that the proposed algorithm has satisfactory tracking performance in the situations where the sensor measurements are biased.