Multi-hypothesis Tracking Research Articles

Multi-Hypothesis Tracking in a Graph-Based World Model for Knowledge-Driven Active Perception

Multi-Hypothesis Tracking in a Graph-Based World Model for Knowledge-Driven Active Perception

SA-HPMHT for Maritime Dim Targets Tracking With Sensor Location Uncertainty

To address the problem of tracking low signal-to-clutter ratio (SCR) maritime targets with sensor location uncertainty, an effective track-before-detect (TBD) method, called the state augmentation histogram probabilistic multihypothesis tracking (SA-HPMHT) algorithm, is proposed in this article. In this algorithm, the initialization of new tracks is realized based on the constant false alarm rate (CFAR) detector with a low threshold and the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${K}$ </tex-math></inline-formula> -means clustering algorithm. The target and sensor states are then simultaneously estimated in an iterative manner via a synthetic histogram and state augmentation (SA) method. In each iteration, we update the estimated sensor state and yield more accurate measured images, allowing for improved performance of detection and tracking targets. Finally, an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${M}/{N}$ </tex-math></inline-formula> logic technology is invoked to confirm or terminate the tracks. The results of simulation and real dataset verification show that the SA-HPMHT algorithm has a better performance compared with the histogram probabilistic multihypothesis tracking (H-PMHT) algorithm and the sequential Monte Carlo probability hypothesis density (SMC-PHD) algorithm.

Time Slotted Multiple-Hypothesis Interference Tracking in Wireless Networks

Nowadays Internet of Things and Industry 4.0 devices are often connected wirelessly. Current wireless sensor network (WSN) deployments are relying in most cases on the industrial, scientific, and medical (ISM) bands without centralized resource scheduling. Thus, each device is a potential source of interference to other devices, both within its own WSN but also to devices in other collocated WSNs. If the transmission behavior of devices from other WSNs is not random, we are able to find patterns in the time domain in their channel access. This is, for example, possible for periodic channel access, which is quite common for WSNs with demanding low-power and reliability requirements. The main goal of this work is to detect multiple sources of periodic interference in time slotted signal-level measurements and estimate the time windows of future transmissions. This gives a WSN a certain understanding of the radio surrounding and can be used to adapt the transmission behavior to thus avoid collisions. For this, the multihypothesis tracking algorithm is adapted and used together with timeslot-based interference measurements on low-cost sensor nodes. The applicability of the algorithm is shown with extensive simulations and the performance is demonstrated with measurements on a time-division multiple access-based WSN built upon the Bluetooth low-energy physical layer.

Apple orchard production estimation using deep learning strategies: A comparison of tracking-by-detection algorithms

The automated detection and counting of fruit in tree canopies is a key component of yield estimation systems, which are indispensable for the precision management of modern orchards. Detection and counting tasks in agricultural environments are not trivial because of challenges such as characteristics of the tree canopies, occlusion caused by leaves and the lighting conditions, among other factors. With the aim of identifying which algorithm is more suitable for yield estimation, we present a comprehensive comparison of tracking-by-detection algorithms, applied to apple counting. The tracking strategies evaluated were Kalman Filter, Kernelized Correlation Filter, Simple Online Real-Time Tracking, Multi Hypothesis Tracking, and Deep Simple Online Real-Time Tracking. The five tracking algorithms were further assessed on two novel databases constructed for this research in Multiple Object Tracking MOT format. After a sensitivity analysis of the trackers, the results show that the most robust approach is the Multiple Hypothesis Tracking, followed by the Deep Simple Online Realtime (DeepSORT), with a MOT accuracy of 97.00% and 93.00%, respectively, when having perfect detection. However, in an application case including a deep learning-based detection stage, the DeepSORT tracker obtains the lowest counting error, which on average for all videos is 20.07% and 31.52% when using YoloV5 and Faster R-CNN as detection strategies. Statistically similar results were obtained using the Kalman Filter with a counting error of 20.5% and 31.9% when detecting fruit with YoloV5 and Faster R-CNN.

Multi-cue multi-hypothesis tracking with re-identification for multi-object tracking

Multi-cue multi-hypothesis tracking with re-identification for multi-object tracking

Convex Variational Inference for Multi-Hypothesis Fractional Belief Propagation Based Data Association in Multiple Target Tracking Systems

The success of Multi-hypothesis tracking(MHT) lies in the use of multiple scans, which can often yield improved tracking performance over single scan based data association methods in radar-centric multi-target tracking(MTT) systems. However, with the increase of the number of targets associated with measurements, there is an exponentially increasing need for formulating potential hypotheses, of which the computational cost may be prohibitively expensive. In this paper, a multi-hypothesis fractional belief propagation (MHFBP) based data association algorithm is proposed by the use of a probabilistic graph model for both the previous trajectories and the current measurements. To achieve this idea, there are two main steps. First, the trajectory-related indicators associated state variables are created by making use of a convex fractional free energy (FFE)function. Second, the convex optimization algorithm is used for the objective function and the fractional belief propagation (FBP) is exploited to obtain the best marginal belief of measurement for target association. In this way, incorrect hypotheses with extremely low probability can be eliminated. Finally, the proposed method is applied to multiple scenarios for MTT by indoor radar system. From the results, we can observe that it provides higher tracking performance compared with the classical MHT, feature-aided MHT(FA-MHT) and MHT-belief propagation (MHT-BP). In addition, we see that the computational burden of the proposed method is reduced significantly to avoid the phenomenon of exponential growth with increasing the number of targets.

Improved N - Best Multiple Hypothesis Tracking Algorithm

In the tracking process of the multi-hypothesis tracking algorithm, many hypothesis are made about the correlation between the track and the measured values. Although its tracking performance is better than other tracking algorithms, it needs a lot of computation time and memory resources when tracking multiple targets or in environments with large clutter due to exponential growth of hypothesis. In view of this situation, an improved N-Best algorithm is proposed to be applied to MHT. The simulation results show that this method can reduce the computation time and realize the track correlation, so it has certain practicability.

A Tale of Three Priors

The histogram-probabilistic multihypothesis tracker (H-PMHT) is an efficient, track-before-detect algorithm that can operate directly on detection surfaces such as data from a focal-plane array (FPA). It provides good performance for low computational cost but is sensitive to sensor parameter changes, e.g. number of channels, gain, normalization, etc. This is because H-PMHT interprets the measured data as a histogram of point measurements; this sensor model is analogous to an FPA counting the number of photons that fall within each of its pixels. Typically this photon count is enormous, which can overwhelm the prior distribution describing target motion. The H-PMHT algorithm addresses this issue by applying a target's prior distribution to every measurement. For a linear, Gaussian target, this results in a data-dependent scaling term on both the measurement and the process noise covariance matrices. However, if the size or gain of an FPA is changed, the photon count changes significantly, and these scaling terms can cause unpredictable performance. One solution to this problem is to change the H-PMHT so that a target's prior distribution is only applied to photons from the same target. This article describes two different approaches to constructing target prior distributions that satisfy this constraint. The benefits of adopting either of these new target prior distributions are demonstrated using simulation.

Dual-Branch CNNs for Vehicle Detection and Tracking on LiDAR Data

We present a novel vehicle detection and tracking system that works solely on 3D LiDAR information. Our approach segments vehicles using a dual-view representation of the 3D LiDAR point cloud on two independently trained convolutional neural networks, one for each view. A bounding box growing algorithm is applied to the fused output of the networks to properly enclose the segmented vehicles. Bounding boxes are grown using a probabilistic method that takes into account also occluded areas. The final vehicle bounding boxes act as observations for a multi-hypothesis tracking system which allows to estimate the position and velocity of the observed vehicles. We thoroughly evaluate our system on the KITTI benchmarks both for detection and tracking separately and show that our dual-branch classifier consistently outperforms previous single-branch approaches, improving or directly competing to other state of the art LiDAR-based methods.

Inventory Tracking for Unstructured Environments via Probabilistic Reasoning

Workpiece location is critical to efficiently plan actions downstream in manufacturing processes. In labor-intensive heavy industries, like construction and shipbuilding, multiple stakeholders interact, stack and move workpieces in the absence of any system to log such actions. While track-by-detection approaches rely on sensing technologies such as Radio Frequency Identification (RFID) and Global Positioning System (GPS), cluttered environments and stacks of workpieces pose several limitations to their adaptation. These challenges limit the usage of such technology to presenting the last known position of a workpiece with no further guidance on a search strategy. In this work we show that a multi-hypothesis tracking approach that models human reasoning can provide a search strategy based on available observations of a workpiece. We show that inventory tracking problems under uncertainty can be approached like probabilistic inference approaches in localization to detect, estimate and update the belief of the workpiece locations. We present a practical Internet-of-Things (IoT) framework for information collection over which we build our reasoning. We also present the ability of our system to accommodate additional constraints to prune search locations. Finally, in our experiments we show that our approach can provide a significant reduction against the conventional search for missing workpieces, of up to 80% in workpieces to visit and 60% in distance traveled. In our experiments we highlight the critical nature of identifying stacking events and inferring locations using reasoning to aid searches even when direct observation of a workpiece is not available.

Improved Active Sonar Tracking in Clutter Using Integrated Feature Data

Target tracking has traditionally focused on the use of kinematic information, such as bearing, range, position, or speed. In some cases, it is possible to extract additional target information that may be useful in the maintenance of reliable tracks, or the establishment of target type or identity. Several methods have been proposed for the use of feature information in target tracking. In this paper, we discuss and evaluate its use as a component of the cumulative likelihood ratio used for track management and contact assignment in a track-oriented multihypothesis tracker (MHT) using a sorted branch and bound approach to contact assignment. The active sonar sea trial data used in this evaluation, Clutter09, include feature values based on the perceptual capabilities of the human auditory system. Three measurement scenarios are considered. Comparisons between the baseline case, which used only kinematic information, and several feature-aided cases showed that the performance of the tracker improved noticeably when feature information was used. The MHT, which was working at a depth of six scans, was also able to tolerate seven times as many false contacts in the feature-aided cases as in the baseline cases. This corresponds to improved detection and tracking of targets that are weaker, or at greater range, or in a more cluttered environment. Results were also significantly better when compared with cases where the feature information was used to prefilter the contacts entering the tracker or to postfilter the tracks produced by the tracker.

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.

Bayesian approach to multiple extended targets tracking with random hypersurface models

A Bayesian approach to multiple moving extended targets tracking is proposed for estimating the shape approximation of the extended targets in addition to their kinematics. Within this approach, the extended target extensions are modelled with random hypersurface models, and a new variant of probabilistic multi-hypothesis tracking is used for modelling assignments of measurements to extended targets. Moreover, an approximate measurement update that arises directly from the analytical techniques of the variational Bayesian framework is derived to simultaneously estimate the posterior states iteratively including the shape and kinematics of each extended target. The performance of the proposed algorithm is demonstrated with simulated data.

DA-PMHT for Multistatic Passive Radar Multitarget Tracking in Dense Clutter Environment

Multistatic passive radar exploits illuminators of opportunity such as radio or television stations for tracking airborne targets. This paper focuses on broadcast signals in a single frequency network modulated according to the digital audio/video broadcasting (DAB/DVB) standards. The main challenge is that one has to solve the measurement-to-illuminator association ambiguity in addition to the conventional measurement-to-target association ambiguity, which increases data association complexity greatly. In this paper, to solve the additional measurement-to-illuminator data association ambiguity, a novel deterministic annealing probabilistic multi-hypothesis tracker based on the extended Kalman filter is proposed. The proposed algorithm works directly in the original three-dimensional Cartesian space, and can efficiently handle a large number of association hypotheses among the targets, measurements, and illuminators in a natural way, also it is computationally attractive. As the angular measurement is not available or often of extremely poor quality, this paper assumes measurements of bistatic range and range rate and does not process bearing information. The simulation results show that the proposed algorithm works reasonably well in dense clutter environment for multistatic passive radar system.

Parameter state estimation for bistatic sonar systems

The use of autonomous underwater vehicles (AUVs) cooperating in a network for anti-submarine warfare surveillance operations is of topical interest. Each AUV has to localise and track targets (submarines) robustly and precisely. This can be realised by a bistatic sonar configuration. The precise knowledge of the bistatic system parameters is mandatory for target tracking. These are in particular the positions of the acoustic sources, transmission times, and the position and heading of the AUV sonar sensor. However, these parameters often are not precisely known, e.g. for non-cooperative acoustic sources or due to navigation uncertainties of the AUV. Therefore, the authors consider the inverse problem, i.e. they estimate the bistatic system parameters by exploiting sonar echoes from known stationary `targets' like wrecks or small islands. Their implementation of the estimation method is based on the multihypothesis tracking technique. Results are discussed for two applications: The first one is estimation of the parameters of a non-cooperative source. The second application focuses on the estimation of the receiver parameters; in particular they show that their approach can be used to increase robustness of AUV navigation. Their algorithms are tested with simulated and real data recorded by the Centre for Maritime Research and Experimentation.

Reducing Processing Time for Histogram PMHT Algorithm in Video Object Tracking

DOI: 10.26650/electrica.2018.36247 This paper describes a novel approach for reducing the processing time of the histogram probabilistic multi-hypothesis tracker (H-PMHT) algorithm in video applications. Video data of flying vehicles is taken from surface to air, and a temporal difference-based technique is applied to video frames for meeting the intensity demands of H-PMHT algorithm. This technique also enables discrimination between objects and eliminates clutter. Variations between the structures of the standard and the improved version of H-PMHT algorithms are described. In addition, the improved H-PMPT is compared with the standard H-PMHT and another approved tracking algorithm to evaluate the performance and processing time reduction ratings.

Multipath Maximum Likelihood Probabilistic Multihypothesis Tracker for Low Observable Targets

In many practical scenarios with multipath propagation, one target may generate multiple detections in one scan. Proper use of multipath-induced measurements can improve the detection of very low observable (VLO) targets. In this paper, a true multitarget tracker, the joint multipath maximum likelihood probabilistic multihypothesis tracker (JMP-ML-PMHT) is proposed to address this problem. The standard ML-PMHT is extended to incorporate multipath detections and jointly track multiple VLO targets. The Cramer–Rao lower bound with multipath detections is derived. Simulation results with an over-the-horizon-radar scenario show that the JMP-ML-PMHT can detect and track multiple VLO targets by effectively utilizing the information in multipath measurements.

Automatic MHT Method Based on Target Characteristics Applied in the Small Space Targets

The Multi-Hypothesis Tracking (MHT) method is one of the most promising approaches to realize automatic tracking of small and dim targets in complex space scene. However, the number of hypotheses generated by this algorithm has an exponentially increasing relationship with the number of trajectories, measurement data, and the number of image frames. In the paper, the automatic detection and tracking method of judging the confidence coefficient of the trajectory obtained by the MHT which is based on the characteristics of dim targets, has been realized in the IRST system. The performance of this method is greatly improved.

Multi-hypothesis tracking of the tongue surface in ultrasound video recordings of normal and impaired speech.

Characterizing tongue shape and motion, as they appear in real-time ultrasound (US) images, is of interest to the study of healthy and impaired speech production. Quantitative anlaysis of tongue shape and motion requires that the tongue surface be extracted in each frame of US speech recordings. While the literature proposes several automated methods for this purpose, these either require large or very well matched training sets, or lack robustness in the presence of rapid tongue motion. This paper presents a new robust method for tongue tracking in US images that combines simple tongue shape and motion models derived from a small training data set with a highly flexible active contour (snake) representation and maintains multiple possible hypotheses as to the correct tongue contour via a particle filtering algorithm. The method was tested on a database of large free speech recordings from healthy and impaired speakers and its accuracy was measured against the manual segmentations obtained for every image in the database. The proposed method achieved mean sum of distances errors of 1.69 ± 1.10mm, and its accuracy was not highly sensitive to training set composition. Furthermore, the proposed method showed improved accuracy, both in terms of mean sum of distances error and in terms of linguistically meaningful shape indices, compared to the three publicly available tongue tracking software packages Edgetrak, TongueTrack and Autotrace.

Detecting Low SNR Tracks With OTHR Using a Refraction Model

Over-the-horizon radar is used to detect targets at very long ranges by reflecting of/refracting signals within the ionosphere. Multiple propagation paths exist between the radar and the target, which can result in multiple target-originated measurements in each scan. It is uncertain which path corresponds with each measurement. Due to the necessary low detection threshold for very low signal-to-noise ratio targets, there are also many false alarms. The maximum-likelihood probabilistic multihypothesis tracker, which is a deep track extractor, is generalized to allow multiple propagation paths. A ray-tracing algorithm is used to model the refraction in the ionosphere. Monte Carlo simulation results are presented for three types of targets: a surface target, a constant altitude target, and a constant vertical acceleration target. We also calculate the Cramer–Rao lower bound for these scenarios, and an analysis is done to determine the probability of track detection, $P_{\text{DT}}$ , for specified probabilities of false track $P_{\text{FT}}$ .