Tracking for drone applications
We are developing efficient visual tracking algorithms suitable for drone tracking applications. A particular quality of these algorithms is that they are able to track the object over a long period of time. They have special mechanisms for detecting the target is absent (e.g., during occlusion or when the target leaves the field of view) and have capabilities to re-detect the target once it reappears. In addition, our tracking algorithms are designed to allow (near) real-time tracking without GPUs.
Results of the most recent research (2018)
Drone tracking a person walking the stairs
(Jon Muhovič, Visual object tracking with a quadcopter, master’s thesis, 2017)
Drone tracking a person in a gym
(Alan Lukežič, Computer-vision-based autonomous control for quadrocopter, Bachelor’s thesis, 2012)