Tracking in dense forest with commercial drones

- EN - DE
Drone Research Professor Bimber
Drone Research Professor Bimber

New method enables people and game tracking in dense forest cover with commercially available drones. Back in the spring, JKU reported on the world’s first real-time tracking of people in dense forest cover. Now comes the next step.

The system can now also be used for normal, commercially available drones. Originally, a special drone with a ten-meter boom to which ten individual cameras were attached was used for this purpose. The imaging technique developed at the JKU (Institute of Computer Graphics; Head: Univ.-Prof. Oliver Bimber), Airborne Optical Sectioning (AOS), allows the occlusion caused by vegetation to be calculated away from the individual images in real time. This provides a clear view of moving objects located on the forest floor. In a next development step, it has now been possible to make this possible for commercially available drones as well. A special drone is thus no longer necessary.

The idea for the new method, Inverse Airborne Optical Sectioning (IAOS), comes from radar technology. In this process, the images taken by a drone hovering stationary above the forest are computationally shifted and integrated. The necessary image shift results from the direction and speed of movement of the object to be tracked on the forest floor. However, since these motion parameters are unknown, JKU researchers have presented a new method to estimate the parameters from the image data despite the occlusion. The result is an occlusion-free view of the searched object (e.g., a missing person, game, or a vehicle). In addition, this object can be subsequently tracked.

Possible applications for the method include missing person searches, game counting and terrain monitoring, e.g. in the protection of critical infrastructure. In the future, AOS should also be able to be used for forest firefighting. Early detection of fire sources, even before visible smoke or fire develops, can provide crucial early detection benefits. The Linz researchers are also working on an extension to drone swarms, in which several drones search in a self-coordinated manner.

Pre-print for publication in Drones Journal:­g/7kM/8UM/­Video2.mp4­g/7kM/8UM/­Video3.mp4