Abstract:

In this presentation I will discuss target tracking problems that we have encountered while collaborating with biologists. The purpose has been to simplify the biologists everyday work, by automating otherwise time consuming manual labor.  Target tracking is generally considered a mature research field; with well established statistical methods that can be used more or less out of the box.  However, the majority of the available literature deals with surveillance scenarios where vehicles or aircraft are tracked.  Tracking of animals and other natural phenomena, which often is the objective for biologists, is less well covered in the standard tracking literature and result in new challenges.  Examples will be presented of how to extended standard methods to better handle the encountered problem formulations; eg, by introducing a new constrained motion model to be able to follow dolphins in a partially occluded basin, or extracting visual queues to detect the take-off time and direction of migrating birds in Emlen funnels.  Though going slightly outside the standard methods, these tricks fit nicely into the general tracking framework.

Biosketch:

Gustaf Hendeby is Associate Professor and Docent in Automatic Control, in the Department of Electrical Engineering, Linköping University.  He received his MSc in Applied Physics and Electrical Engineering in 2002 and his PhD in Automatic Control in 2008, both from Linköping University.  He worked as Senior Researcher at the German Research Center for Artificial Intelligence (DFKI) 2009-2011, and Senior Scientist at Swedish Defense Research Agency (FOI) and held an adjunct Associate Professor position at Linköping University 2011-2015. Dr. Hendeby's main research interests are stochastic signal processing and sensor fusion with applications to nonlinear problems, target tracking, and simultaneous localization and mapping (SLAM).  He has experience of both theoretical analysis as well as implementation aspects.

Gustaf Hendeby's homepage