Tracking of non-brownian particles using the Viterbi algorithm
2015 (English)In: Proceedings - International Symposium on Biomedical Imaging, IEEE conference proceedings, 2015, 380-384 p.Conference paper (Refereed)Text
We present a global tracking algorithm for tracking particles with dynamic motion models. The tracking algorithm augments a existing global track linking algorithm based on the Viterbi algorithm with a Gaussian Mixture Probability Hypothesis Density filter. This allows the tracking algorithm to use the target velocities to link tracks. The algorithm can handle clutter, missed detections, and random appearance and disappearance of particles in the field of view. The algorithm can also handle targets that switch between different motion models according to a Markov process. The algorithm is evaluated on the synthetic datasets used in the ISBI 2012 Particle Tracking Challenge, which simulate vesicles, receptors, microtubules, and viruses at different particle densities and signal to noise ratios. The evaluation shows that our algorithm performs well across a wide range of particle tracking problems in both 2D and 3D.
Place, publisher, year, edition, pages
IEEE conference proceedings, 2015. 380-384 p.
GM-PHD, Particle tracking, Viterbi algorithm, Markov processes, Medical imaging, Motion analysis, Probability density function, Signal to noise ratio, Target tracking, Tracking (position), Viruses, Dynamic motion models, Gaussian mixture probability hypothesis density filters, Linking algorithms, Non-brownian particles, Synthetic datasets, Tracking algorithm, Algorithms
IdentifiersURN: urn:nbn:se:kth:diva-181653DOI: 10.1109/ISBI.2015.7163892ISI: 000380546000091ScopusID: 2-s2.0-84944319065ISBN: 9781479923748OAI: oai:DiVA.org:kth-181653DiVA: diva2:912437
12th IEEE International Symposium on Biomedical Imaging, ISBI 2015, 16 April 2015 through 19 April 2015
QC 201603162016-03-162016-02-022016-09-05Bibliographically approved