An algorithm for tracking swimming worms (bibtex)
by Carmine Sansone
Abstract:
During spawning, the marine worms Platynereis dumerilii exhibit certain swimming be- haviours, which are described as a 'nuptial dance'. We propose a 2D tracking approach that enables the extraction of data to quantify behaviours. With Platynereis dumerilii worms, associations based on appearance are challenging since they are deformable and regularly change shape and luminosity depending on their movement. During the spawning phase, the worms exhibit complex interactions leading to occlusions which interrupt the continuous track of a worm. During the occlusion the worms overlap themselves several times and change the movement direction. To maintain the individual identities within a video sequence, we propose an approach where appearance models are used to compare a set of feature before and after an occlusion occurs to get correct associations of worm identities. The chosen features are: the normalized shape (which is computed using a novel approach), the area, the length, the luminosity and the position. To correctly de ne the head and the tail points of the worms a trajectory analysis is done and depending on the movement of the worms the head and the tail points are detected. The method that we propose can be easily re-adapted to track similar type of worms. The evaluation done shows how promising is this new approach for the tracking of two Platynereis dumerilii worms.
Reference:
An algorithm for tracking swimming worms (Carmine Sansone), Technical report, PRIP, TU Wien, 2016.
Bibtex Entry:
@TechReport{TR139,
  author =	 "Carmine Sansone",
  title =	 "An algorithm for tracking swimming worms",
  institution =	 "PRIP, TU Wien",
  number =	 "PRIP-TR-139",
  year =	 "2016",
  url =		 "ftp://ftp.prip.tuwien.ac.at/pub/publications/trs/tr139.pdf",
  abstract =	 "During spawning, the marine worms Platynereis dumerilii exhibit certain swimming be-
haviours, which are described as a 'nuptial dance'. We propose a 2D tracking approach
that enables the extraction of data to quantify behaviours. With Platynereis dumerilii
worms, associations based on appearance are challenging since they are deformable and
regularly change shape and luminosity depending on their movement. During the spawning
phase, the worms exhibit complex interactions leading to occlusions which interrupt the
continuous track of a worm. During the occlusion the worms overlap themselves several
times and change the movement direction. To maintain the individual identities within a
video sequence, we propose an approach where appearance models are used to compare
a set of feature before and after an occlusion occurs to get correct associations of worm
identities. The chosen features are: the normalized shape (which is computed using a novel
approach), the area, the length, the luminosity and the position. To correctly dene the
head and the tail points of the worms a trajectory analysis is done and depending on the
movement of the worms the head and the tail points are detected.
The method that we propose can be easily re-adapted to track similar type of worms.
The evaluation done shows how promising is this new approach for the tracking of two
Platynereis dumerilii worms.",
}
Powered by bibtexbrowser