Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

DY: Fachverband Dynamik und Statistische Physik

DY 7: Statistical physics II (general)

DY 7.12: Vortrag

Dienstag, 24. März 2009, 12:45–13:00, HÜL 386

A New Method of Tracking Feature Points in Videos — •Mario Heidernätsch and Günter Radons — Chemnitz University of Technology, D-09126 Chemnitz, Germany

A general problem in the analysis of moving objects is to trace their path in a series of snapshots or a video. The objects are abstractly described by a set of properties/features, e.g. position, shape, color etc. To determine whether a detected feature point is the same in a series of consecutive images and to follow its path, it is necessary to find the optimal assignment by means of its features. An additional difficulty which could show up during tracking lies in the property that points stay undetected on single frames and make it inevitable to look at more than one frame at a single tracking step.

To solve this problem we propose a new tracking algorithm, which extends the idea of a connection cost matrix [1] with use of the Viterbi algorithm. On the basis of random moving fluorescent molecules we demonstrate how the algorithm works. In addition to the minimalistic a priori information of a single particle track due to the random movement, these molecules underlie the effect of photobleaching and therefore are temporary complete invisible. Thus these videos are the best “Testground” for this algorithm.

[1] I.F. Sbalzarini, P. Koumoutsakos, Journal of Structural Biology, 151, pp. 182-195, 2005