Vision-based crop row detection system for UAV-based weed detection in arboriculture

Abstract

Due to legal changes, the use of agents for soil disinfection in arboriculture is severely limited. For this reason, tree nurseries must look for new ways in which weed control can be carried out on seed beds in the future. In cooperation with the local Chamber of Agriculture we want to explore the use of Unmanned Aerial Vehicles (UAVs) for the autonomous weed detection and weed management in arboriculture to develop an innovative approach for weed control. UAVs are selected because of the disadvantage of ground-based robotic systems in agriculture which is the limitation of their usability in wet soils, as most robotic systems get stuck and are no longer manoeuvrable. Climate change is causing these weather conditions to occur more frequently. This paper evaluates early research results of a vision-based crop row detection system for UAV-based weed detection in arboriculture in regards to overall detection performance, real-time capability, weight, size and power-consumption.