Assisted Canopy Sampling Using Unmanned Aerial Vehicles (UAVs)

摘要

Recently, Unmanned Aerial Vehicles (UAVs) carrying canopy sampling tools have been demonstrated. They present serious advantages in terms of safety and reach, but their time efficiency and ease of use could still be improved. Three main challenges were identified when using such tools: the selection of accessible branches, the lack of precision caused by the pendulum movement of the sampling tool, and the necessity to have two operators. In this paper, we address those issues by presenting the foundation of an architecture to provide sampling assistance on deciduous trees that makes UAV sampling more efficient and easy to use. The UAV is equipped with a DeLeaves canopy sampling tool, an RGB-D camera, and an onboard computer. Computer vision was used to detect and localize branches. Sampling assistance was used to limit the tool's oscillation and reduce the pilot's workload, allowing him or her to control both the UAV and the tool. This assisted canopy sampling solution allowed a sampling phase time of 34 seconds on average on mature deciduous trees, thus reducing the time required by 7 times.