Split and merge data association filter for dense multi-target tracking

Genovesio, Auguste; Olivo-Marin, J.-C.

doi:10.1109/icpr.2004.1333863

Cited by 46 publications

(33 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Trajectories can be established by connecting particles from one frame to the next using nearest-neighbour association and the motion history of individual particles (step 3) 13,57,85 . Furthermore, a split-and-merge mechanism allows the continued tracking of particles even when they merge or separate into smaller entities 86 . The overlap in fluorescence emission between two or more coloured channels can be used to probe the assembly or disassembly of different viral components or the interactions between viral and cellular structures.…”

Section: Box 3 | Image Analysismentioning

confidence: 99%

Virus trafficking – learning from single-virus tracking

2007

View full text Add to dashboard Cite

What could be a better way to study virus trafficking than 'miniaturizing oneself' and 'taking a ride with the virus particle' on its journey into the cell? Single-virus tracking in living cells potentially provides us with the means to visualize the virus journey. This approach allows us to follow the fate of individual virus particles and monitor dynamic interactions between viruses and cellular structures, revealing previously unobservable infection steps. The entry, trafficking and egress mechanisms of various animal viruses have been elucidated using this method. The combination of single-virus trafficking with systems approaches and state-of-the-art imaging technologies should prove exciting in the future.

show abstract

Section: Box 3 | Image Analysismentioning

confidence: 99%

Virus trafficking – learning from single-virus tracking

2007

View full text Add to dashboard Cite

show abstract

“…To address interactions in this context, measurement models have been modified to model merged measurements between target. Explicitly modeling these shared measurements has been shown to be more effective tracking through occlusions [14], [26], [5], [27]. This is the approach we use in this work.…”

Section: Related Workmentioning

confidence: 99%

“…Background subtraction algorithms return multiple blob centroids per target and, during close interactions, often return a merged blob centroid for two or more targets [41], [14]. Interest point detectors return a cloud of multiple measurements around a target and, as shown in the example of Fig.…”

Section: Introductionmentioning

confidence: 99%

MCMC Data Association and Sparse Factorization Updating for Real Time Multitarget Tracking with Merged and Multiple Measurements

Khan

Balch

Dellaert

2006

IEEE Trans. Pattern Anal. Mach. Intell.

109

100

View full text Add to dashboard Cite

Abstract-In several multitarget tracking applications, a target may return more than one measurement per target and interacting targets may return multiple merged measurements between targets. Existing algorithms for tracking and data association, initially applied to radar tracking, do not adequately address these types of measurements. Here, we introduce a probabilistic model for interacting targets that addresses both types of measurements simultaneously. We provide an algorithm for approximate inference in this model using a Markov chain Monte Carlo (MCMC)-based auxiliary variable particle filter. We Rao-Blackwellize the Markov chain to eliminate sampling over the continuous state space of the targets. A major contribution of this work is the use of sparse least squares updating and downdating techniques, which significantly reduce the computational cost per iteration of the Markov chain. Also, when combined with a simple heuristic, they enable the algorithm to correctly focus computation on interacting targets. We include experimental results on a challenging simulation sequence. We test the accuracy of the algorithm using two sensor modalities, video, and laser range data. We also show the algorithm exhibits real time performance on a conventional PC.

show abstract

“…Problems with object segmentation often occur (Genovesio 2004;Kumar et al 2006): when another region occludes the object (a fixed object in the scene or other moving object), when the object image is split into fragments during image segmentation, or when the images from different objects get merged because of their closed or overlapped projection on the camera plane. Besides, extraneous elements in the scene such as waving trees, smoke, clouds, etc., may originate false detected regions interacting with the real objects of interest but they should not degrade their continuity.…”

Section: Sensor Data Association and Soft-computing Approachesmentioning

confidence: 99%

“…This limitation was systematically assumed in the first applications to visual data association (Cox 1993;Cox et al 1995), but it can be too restrictive for video processing under situations of occlusions and image splitting. Recent approaches have identified the problem and proposed the extension of previous algorithms to take into account the splitting/merging effects for visual data association (Kumar et al 2006;Genovesio and Olivo-Marin 2004;Liu et al 2009;Rasmussen et al 2001;Sheikh et al 2008). The detected blobs corresponding to each target must be re-connected before they are used to update each track (Genovesio 2004).…”

Section: Sensor Data Association and Soft-computing Approachesmentioning

confidence: 99%

Fuzzy region assignment for visual tracking

et al. 2011

View full text Add to dashboard Cite

In this work we propose a new approach based on fuzzy concepts and heuristic reasoning to deal with the visual data association problem in real time, considering the particular conditions of the visual data segmented from images, and the integration of higher-level information in the tracking process such as trajectory smoothness, consistency of information, and protection against predictable interactions such as overlap/occlusion, etc. The objects' features are estimated from the segmented images using a Bayesian formulation, and the regions assigned to update the tracks are computed through a fuzzy system to integrate all the information. The algorithm is scalable, requiring linear computing resources with respect to the complexity of scenarios, and shows competitive performance with respect to other classical methods in which the number of evaluated alternatives grows exponentially with the number of objects.

show abstract

Split and merge data association filter for dense multi-target tracking

Cited by 46 publications

References 10 publications

Virus trafficking – learning from single-virus tracking

Virus trafficking – learning from single-virus tracking

MCMC Data Association and Sparse Factorization Updating for Real Time Multitarget Tracking with Merged and Multiple Measurements

Fuzzy region assignment for visual tracking

Contact Info

Product

Resources

About