The importance of novel methods for weed control is rapidly growing, however, evaluating the effectiveness of these technologies is lacking. Weed AI is working on methods to rapidly and autonomously evaluate the efficacy of weeding systems. Based on core deep learning (AI) methods for plant recognition, this project will develop vision-based methods to automatically assess the effectiveness of weeding operations (both weed and crop). Funded by the German Federal Ministry of Food and Agriculture (BMEL), it is hoped that this project will contribute to assessing new weeding technologies and improve uptake of these new technologies.
Precision Weed Management
Developing precision weed management systems to enable alternative weed management deployed on a robotic system. These approaches will be applied in a minimally invasive manner to reduce both inputs (e.g herbicides and fuel) and soil disturbance.
Automated Horticulture Crop Surveying
The Agricultural Robotics group is developing an automated Crop Surveying platform. It consists of a retrofitted glasshouse platform, equipped with an extensive sensor suite and a robot arm on top. Driving this project is the development of novel algorithms to exploit the multitude of information available including RGB, depth (stereo), etc. to produce better instance-based semantic segmentation for estimating the state of the crop to drive 4D crop mapping and intervention systems.
Automated Monitoring and Decision Making for High-Value Crops
This pilot project aims to explore how to efficiently plan the motion of the robotic arm (with an eye-in-hand sensor) to best sense the state of the space (the crop and environment). This is a key enabling factor for automatically sensing the crop through its life cycle. This short-term project is funded by the TRA-6 at the University of Bonn and is fostering collaboration between the Agricultural Robotics group and the Humanoid Robots Lab.