The GM-PHD Filter Multiple Target Tracker

Clark, Dan E.; Panta, K.; Vo, Ba-Ngu

doi:10.1109/icif.2006.301809

Cited by 111 publications

(97 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We omit the labelling of the Gaussian components to maintain target continuity [4] for simplicity of presentation. These are summarized below:…”

Section: Gaussian Mixture Implementationsmentioning

confidence: 99%

See 1 more Smart Citation

Gaussian mixture implementations of probability hypothesis density filters for non-linear dynamical models

Clark

et al. 2008

IET Seminar on Target Tracking and Data Fusion: Algorithms and Applications

View full text Add to dashboard Cite

The Probability Hypothesis Density (PHD) filter is a multipletarget filter for recursively estimating the number of targets and their state vectors from sets of observations. The filter is able to operate in environments with false alarms and missed detections. Two distinct algorithmic implementations of this technique have been developed. The first of which, called the Particle PHD filter, requires clustering techniques to provide target state estimates which can lead to inaccurate estimates and is computationally expensive. The second algorithm, called the Gaussian Mixture PHD (GM-PHD) filter does not require clustering algorithms but is restricted to linear-Gaussian target dynamics, since it uses the Kalman filter to estimate the means and covariances of the Gaussians. This article provides a review of Gaussian filtering techniques for non-linear filtering and shows how these can be incorporated within the Gaussian mixture PHD filters. Finally, we show some simulated results of the different variants.

show abstract

“…We omit the labelling of the Gaussian components to maintain target continuity [4] for simplicity of presentation. These are summarized below:…”

Section: Gaussian Mixture Implementationsmentioning

confidence: 99%

“…In the Gaussian mixture implementation, the multiple target states are estimated by taking the Gaussian components with highest weights and tracks can be maintained by labelling the Gaussian components [4]. More complex methods for dealing with target resolution uncertainty have been developed using this approach as a basis [5].…”

Section: Introductionmentioning

confidence: 99%

Gaussian mixture implementations of probability hypothesis density filters for non-linear dynamical models

Clark

et al. 2008

IET Seminar on Target Tracking and Data Fusion: Algorithms and Applications

View full text Add to dashboard Cite

show abstract

“…Since the GM-PHD filter in [VM06] does not ensure the continuity of individual object tracks, it is extended by [CPV06] to include track labels. The convergence analysis of the filter is given in [CV07].…”

Section: The Gaussian-mixture Phd Filtermentioning

confidence: 99%

Intersection-Based Road User Tracking Using a Classifying Multiple-Model PHD Filter

Meißner

Reuter

Strigel

et al. 2014

IEEE Intell. Transport. Syst. Mag.

View full text Add to dashboard Cite

Bibliografische Information der Deutschen NationalbibliothekDie Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.dnb.de abrufbar. AbstractDue to an increasing number of traffic accident fatalities at intersections, the prevention of traffic accidents at intersection attracts more and more attention in public as well as in research. One example therefore, is the joint-project Ko-PER of the research initiative Ko-FAS which aims to increase the preventive traffic safety. This is realized by an preferably exhaustive recognition of the vehicles' environments based on locally gathered and communicated perception data. Therefore, especially at urban intersections, an intersection-based perception system which communicates the current traffic situation to vehicles in the intersection's surrounding is essential. Thus, a busy public intersection in Aschaffenburg, Germany was equipped with an intersection perception system, which contains multiple laserscanners and video cameras as well as data processing units. Beside the planning of the perception system, the core of this thesis is the laserscanner-based detection and classification of road users and the tracking of them by fusing detections of the laserscanners and video cameras.Since the laserscanners at the intersection are mounted at posts of lamps, traffic-lights as well as surrounding buildings, segmentation algorithms are introduced which utilize the knowledge about the consistent environment to segment the laserscanner measurements in background and foreground. Subsequently, the object detection in the foreground data relies on position as well as motion information. The 3D measurement points are clustered with a real-time capable implementation of the density based spatial clustering for applications with noise (DBSCAN) algorithm. Nevertheless, particularly large objects are sometimes not entirely detected due to occlusions, poor reflection characteristics and measurement uncertainties. This may lead to the separation of one object in multiple clusters. To avoid the overclustering, a combination of 2D tracking and 3D clustering methods is proposed. Beside the benefit of the considerably more robust object detection, additionally the determination of the objects' orientations is improved. Finally, the detected objects are classified by means of features of their laserscanner point clusters.The focus of this thesis lays on the tracking of road users at traffic intersections. Thus, a tracking approach is required which is able to handle multiple objects detected by multiple classifying sensors with different field of views. Due to the topdown modeling of the multi-object tracking problem and the availability of efficient implementations, a Gaussian mixture probability hypothesis density (GM-PHD) filter is used. Based on this the classifying multiple-model probability hypothesis density (CMMPHD) filter is developed which facilitates to track road user...

show abstract

“…The labeling methods are operated in the implementation process of PHD/CPHD filtering. In the implementation, the elements used for representing target are particles in SMC implementation [154] or Gaussian element in GM implementation [131,[155][156][157][158]. The labeling methods partition the elements in the position domain and give each element in the same partition with the same label.…”

Section: ) Target Birth and Spawning Modelmentioning

confidence: 99%

A Survey of Motion-Based Multitarget Tracking Methods

Qiu¹,

Zhang²,

Lü³

et al. 2015

PIER B

View full text Add to dashboard Cite

Abstract-Multitarget tracking (MTT) in surveillance system is extremely challenging, due to uncertain data association, maneuverable target motion, dense clutter disturbance, and real-time processing requirements. A good many methods have been proposed to cope with these challenges. However, no up-to-date survey is available in the literature that can help to select suitable tracking algorithm for practical problem. This paper provides a comprehensive review of the state-of-the-art motion-based MTT techniques, classifies existing methods into two groups, i.e., the detect-beforetrack (DBT) scheme and the track-before-detect (TBD) scheme. The DBT scheme is employed to achieve robust and tractable tracking performance when the signal-noise-ratio (SNR) is strong. The TBD scheme is used in the scenarios of low SNR, and it aims to cumulate target energy by multiple sensor frames. Furthermore, depending on the data association mechanism, the DBT methods can be classified into two categories, data association based approaches and finite set statistics (FISST) based approaches. And the TBD methods can be classified into non-Bayesian approaches and Bayesian approaches depending on the basis theory used for tracking. For each category, this paper provides the detailed descriptions of the representative algorithms, and examines their pros and cons. Finally, new trends for future research directions are offered.

show abstract

The GM-PHD Filter Multiple Target Tracker

Cited by 111 publications

References 12 publications

Gaussian mixture implementations of probability hypothesis density filters for non-linear dynamical models

Gaussian mixture implementations of probability hypothesis density filters for non-linear dynamical models

Intersection-Based Road User Tracking Using a Classifying Multiple-Model PHD Filter

A Survey of Motion-Based Multitarget Tracking Methods

Contact Info

Product

Resources

About