Multi-Target Tracking and Occlusion Handling With Learned Variational Bayesian Clusters and a Social Force Model

ur-Rehman, Ata; Naqvi, Syed Mohsen; Mihaylova, Lyudmila; Chambers, Jonathon A.

doi:10.1109/tsp.2015.2504340

Cited by 47 publications

(25 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…with its process noise covariance Q k−1 = Diag( [25,25,16,16,4,4]), where I 2 and 0 2 are the 2 × 2 identity and zero matrices, and ∆k represents the time interval between frame k and k + 1. The observation model is H k = [I 2 , 0 2 , 0 2 ; 0 2 , 0 2 , I 2 ] and its observation noise covariance is R k = Diag( [25,25]) [2].…”

Section: Implementation Detailsmentioning

confidence: 99%

See 1 more Smart Citation

Multi-Level Cooperative Fusion of GM-PHD Filters for Online Multiple Human Tracking

Angelini

Chambers

et al. 2019

IEEE Trans. Multimedia

Self Cite

View full text Add to dashboard Cite

In this paper, we propose a multi-level cooperative fusion approach to address the online multiple human tracking problem in a Gaussian Mixture Probability Hypothesis Density (GM-PHD) filter framework. The proposed fusion approach consists essentially of three steps. Firstly, we integrate two human detectors with different characteristics (full-body and bodyparts), and investigate their complementary benefits for tracking multiple targets. For each detector domain, we then propose a novel Discriminative Correlation Matching (DCM) model, and fuse it with spatio-temporal information to address ambiguous identity association in the GM-PHD filter. Finally, we develop a robust fusion center with virtual and real zones to make a global decision based on preliminary candidate targets generated by each detector. This center also mitigates the sensitivity of missed detections in the Generalized Covariance Intersection (GCI) fusion process, thereby improving the fusion performance and tracking consistency. Experiments on the MOTChallenge Benchmark demonstrate the proposed method achieves improved performance over other state-of-the-art RFS based tracking methods.

show abstract

Section: Implementation Detailsmentioning

confidence: 99%

“…Offline tracking approaches [8], [10], [12] employ both past and future detections to globally formulate an optimization problem, which is unsuitably applied in real world applications. Online tracking approaches [4], [5], [7], [9], [11], [13] achieve the tracking estimates only relying on detections from past and current time.…”

Section: Introductionmentioning

confidence: 99%

Multi-Level Cooperative Fusion of GM-PHD Filters for Online Multiple Human Tracking

Angelini

Chambers

et al. 2019

IEEE Trans. Multimedia

Self Cite

View full text Add to dashboard Cite

show abstract

“…Many Bayesian filtering algorithms have been developed to tackle the visual multi-object tracking problem, such as particle filter [10,28], joint probabilistic data association filter (JPDAF) [29,30], Marcov chain Mote Carlo (MCMC) data association [31,32], track linking [33][34][35], multiple hypothesis tracking (MHT) [36], kernel based Bayesian filter [37], and Bayesian filters with Relative Motion Network (RMN) [38].…”

Section: Introductionmentioning

confidence: 99%

On-Line Visual Tracking with Occlusion Handling

Rathnayake

Gostar

Hoseinnezhad

et al. 2020

Sensors

View full text Add to dashboard Cite

One of the core challenges in visual multi-target tracking is occlusion. This is especially important in applications such as video surveillance and sports analytics. While offline batch processing algorithms can utilise future measurements to handle occlusion effectively, online algorithms have to rely on current and past measurements only. As such, it is markedly more challenging to handle occlusion in online applications. To address this problem, we propagate information over time in a way that it generates a sense of déjà vu when similar visual and motion features are observed. To achieve this, we extend the Generalized Labeled Multi-Bernoulli (GLMB) filter, originally designed for tracking point-sized targets, to be used in visual multi-target tracking. The proposed algorithm includes a novel false alarm detection/removal and label recovery methods capable of reliably recovering tracks that are even lost for a substantial period of time. We compare the performance of the proposed method with the state-of-the-art methods in challenging datasets using standard visual tracking metrics. Our comparisons show that the proposed method performs favourably compared to the state-of-the-art methods, particularly in terms of ID switches and fragmentation metrics which signifies occlusion.

show abstract

“…In this approach, an attractor point is assigned according to the prior distribution to form an attractive force on the particles. In an analogous way, incorporating the static context knowledge with the particle filter also can be found in visual tracking applications [20], [21].…”

Section: Introductionmentioning

confidence: 99%

Ground Vehicle Tracking Using Context-Based Sojourn Time Dependent Markov Model and Pseudo-Measurement

Tian

Cen

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Tracking maneuvering vehicles in complex dynamic environment is a challenging problem for advanced driver assistance system and autonomous driving systems. Most conventional vehicle tracking algorithms can not model the vehicle dynamic exactly because of the uncertain moving behavior. However, due to the on-road capability, vehicles have to subject to various constraints imposed by traffic rules and roads. Taking advantage of those context information can refine and improve the performance of tracking as it provides additional prior information for vehicles' dynamic behavior. To achieve this goal, this paper presents a novel context-enhanced tracking approach that exploits the context information to reduce the uncertainty of dynamic estimation. A new context-based sojourn-time dependent semi-Markov (STDM) model, called sojourn-time dependent semi-Markov variable structure interacting multiple model (STDM-VSIMM), is proposed to describe the vehicle's longitudinal acceleration process. In order to cope with the context information into STDM model, a context-based Bayesian network is presented to replace the fixed model transition probabilities with sojourn-time dependent transition probabilities. Compared with traditional interacting multiple model tracking with fixed transition probability, this adaptation switching strategy makes the vehicle motion sequence closer to the natural behavior and improve the tracking performance. Furthermore, a novel pseudo-measurement is constructed to formulate the road-map constraint in tracking process for reducing uncertain on mobility constraints. Simulation results shows that the proposed STDM-VSIMM can achieve better performance after considering the context. INDEX TERMS Vehicle tracking, context-enhanced, sojourn-time dependent Markov model, pseudomeasurements.

show abstract

Multi-Target Tracking and Occlusion Handling With Learned Variational Bayesian Clusters and a Social Force Model

Cited by 47 publications

References 33 publications

Multi-Level Cooperative Fusion of GM-PHD Filters for Online Multiple Human Tracking

Multi-Level Cooperative Fusion of GM-PHD Filters for Online Multiple Human Tracking

On-Line Visual Tracking with Occlusion Handling

Ground Vehicle Tracking Using Context-Based Sojourn Time Dependent Markov Model and Pseudo-Measurement

Contact Info

Product

Resources

About