A new automatic video multitracking system able to follow a large school
J. Delcourt1, Ch. Becco2, N. Vandewalle2 and P. Poncin1
1Behavioural Biology Unit University
of Liège, Belgium
2Group for Research and Applications in Statistical Physics
- University of Liège, Belgium
The study of shoaling behaviours is tackled by scientists by three approaches:
the field, laboratory and computer modelling approach. The field studies
are complicated and permit rarely an analysis of individual positions.
The third approach, consisting of development of mathematic models, is
artificial. It requires a comparison with empiric results to validate
and to determine the explications degree of each model. Thanks to
the developments of digital imaging techniques, the laboratory studies
on fish are able to measure the behavioural parameters with precision,
impossible to realize with manual recording.
Multitracking systems are rare and of recent development. These are insufficient
to study the real size of shoals, often composed of hundreds of .shes.
We have developed a new system of multitracking thanks to collaboration
with the G.R.A.S.P. The G.R.A.S.P. has before realised other multitracking
systems to follow some hundreds physical objects. Our video multitracking
system is able to detect and to track more than a hundred unmarked fishes
by gray scaling technique during some minutes. It permits an analysis
as well at the group level as at the individual level. The multitraking
program is able to attribute a number at each fish and to follow each
one during the whole duration of the track. Our system permits the analysis
of the movement of each individual, even if the trajectories of two fishes
cross each other. It is possible thanks to the theoretical estimation
of trajectory of each fish, compared with the real trajectory (analysis
with feedback).
The human eye is always necessary to detect the errors in identification
but reediting is possible. However, if the fishes dont move too
rapidly, a rate of 25 images /s gives very good results, without too much
visual expertise from the experimenter. If the fishes are too fast (i.e.
escape response), it is better to use a high speed camera.
Paper presented
at Measuring Behavior 2005
, 5th International Conference on Methods and Techniques
in Behavioral Research, 30 August - 2 September 2005, Wageningen, The
Netherlands.
© 2005 Noldus
Information Technology bv
|