A method for resolving occlusions when multitracking individuals in a shoal.

Abstract

Studying the collective behavior of fishes often requires tracking a great number of individuals. When many fishes move together, it is common for individuals to move so close to each other that some fishes superimpose themselves on others during one or several units of time, which impacts on tracking accuracy (i.e., loss of fish trajectories, interchange of fish identities). Type 1 occlusions arise when two fishes swim so near each other that they look like one long fish, whereas type 2 occlusions occur when the fishes' trajectories cross to create a T- or X-shaped individual. We propose an image processing method for resolving these types of occlusions when multitracking shoals in two dimensions. We assessed processing effectiveness after videorecording shoals of 20 and 40 individuals of two species that exhibit different shoal styles: zebrafish (Danio rerio) and black neon tetras (Hyphessobrycon herbertaxelrodi). Results show that, although the number of occlusions depended on both the number of individuals and the species, the method is able to effectively resolve a great deal of occlusions, irrespective of the species and the number of individuals. It also produces images that can be used in a multitracking system to detect individual fish trajectories. Compared to other methods, our approach makes it possible to study shoals with water depths similar to those seen in the natural conditions of the two species studied.