LocusTracker: Markerless Multi-Agent Tracking In Various Environments

Roee M. Francos, Moshe Guershon, Amir Ayali, Tali Hatuka, Alfred M. Bruckstein

Abstract:

In biological behavioral experiments precise location and tracking of living organisms is of great importance.
In particular, reliable and accurate video/vision based detection and tracking of mobile agents (animals) in a swarm can
provide valuable information on movement responses, on interaction between individuals, and the influence of the environment’s
geometry and other factors on the emergence of collective behaviour. In this work we present a computer-vision
software tool called the LocusTracker, developed for research on the swarming behaviours of locusts. The tool enables
tracking individual locusts in a group, and was deployed in various environments for several hours, without disturbances
that would be inherent when using physical markers on the moving insects. The tool is an advanced computer-visionbased
multiple blob detection and tracking system that can serve as a useful platform for research on biological and robotic
multi-agent collective behaviours, and hence is released as an open-source software tool for the scientific community.