Content area

Full Text

 
Headnote

Coming from the framework of unmarked fry tracking, we compared the capacities, advantages, and disadvantages of two recent video tracking systems: EthoVision 2.3 and a new prototype of multitracking. The EthoVision system has proved to be impressive for tracking a fry using the detection by gray scaling. Detection by subtraction has given less accurate results. Our video multitracking system is able to detect and track more than 100 unmarked fish by gray scaling technique. It permits an analysis at the group level as well as at the individual level. The multitracking program is able to attribute a number to each fish and to follow each one for the whole duration of the track. Our system permits the analysis of the movement of each individual, even if the trajectories of two fish cross each other. This is possible thanks to the theoretical estimation of the trajectory of each fish, which can be compared with the real trajectory (analysis with feedback). However, the period of the track is limited for our system (about 1 min), whereas EthoVision is able to track for numerous hours. In spite of these limitations, these two systems allow an almost continuous automatic sampling of the movement behaviors during the track.

Scientists study schooling or shoaling behaviors by utilizing three approaches: field studies, computer modeling, and the laboratory approach.

Field studies are complicated and rarely permit an analysis of individual positions. For example, a field study using sonar permits only a global approach with shoals (see, e.g., Axelsen, Anker-Nilssen, Possum, Kvamme, & Nerttestad, 2001; Nerttestad & Axelsen, 1999). The analysis of filmed sequences in nature is very labor intensive and imprecise (see, e.g., Krause, 1993). Radio tracking is only possible for a much smaller number of individuals.

The second approach, consisting of the development of mathematical models by computer scientists (Reynolds, 1987), by biologists (Parrish, Viscido, & Grünbaum, 2002; Suzuki, Tsumonu, & Hiraishi, 2003; Viscido, Parrish, & Grünbaum, 2004), or by physicists (Grégoire & Chaté, 2004; Vicsek, Czirok, Ben-Jacob, Cohen, & Shochet, 1995; Vicsek, Czirók, Farkas, & Helbing, 1999) is artificial. It requires a comparison with empirical results to validate and determine the degree of validity for each model.

Thanks to the development of digital imaging techniques, a third approach in...

Show less