Mitarbeitende

Broad-leaved dock plants are prevalent perennial weeds in meadows and have low nutritional value for ruminants, especially cows. In this study, we explored the potential of unmanned aerial vehicles (UAV) imaging for automated dock recognition on grassland to control these weeds more efficiently. We found that a ground sampling distance of 1.1 mm is needed to distinguish docks with high confidence from other plants with similar visual features (e.g. plan-tain or dandelion), especially under challenging illumination conditions, such as bright sunlight. With our selected UAV system (DJI Matrice 300 RTK with Zenmuse P1 camera) the required image resolution corresponds to flying heights ranging from 12 to 20m using a 50 mm objec-tive covering a swath width of about 8 - 14m. We also introduce our proposed fully automated workflow: this includes uploading images via a 5G cellular network to a server, detecting and pinpointing all docks using an object detection model, followed by a georeferencing mecha-nism resulting in a map that indicates the location of each plant needing treatment. Our re-search will pave the way for the automated management of docks using unmanned vehicles soon.