TC AgRA – Webinar 2 – Robotics Control of Broad Leaved Weeds in Dairy Pastures

Frits K. van Evert received an M.Sc. in Agronomy from Wageningen University and a Ph.D. in Soils from Washington State University. He has worked in general agronomy, analysis of agricultural systems, and vision & robotics. He has worked in Niger, The United States, and The Netherlands. Currently, he is a researcher at Wageningen University and Research Centre and works on nitrogen nutrition and crop growth modeling in potatoes, on sustainability of agricultural systems, and on agricultural robotics.

Abstract
Broad-leaved dock (Rumex obtusifolius L.) is a common and troublesome grassland weed with a wide geographic distribution. In organic farming, the best option to control the weed is manual removal of the plants. We describe a robot to detect and control broad-leaved dock. The robot consists of a diesel-powered frame of 1.25 x 1.11 m with four independently driven wheels. Weeds are detected with a downward-looking camera that provides full-colour images with a resolution of 1.5 mm per pixel. Image processing is based on variance analysis of sub-images (tiles) of 8 x 8 pixels. Weeds are controlled using a cutter with a single 0.20 m blade that rotates around a vertical axis and is pushed into the ground at the location of the weed. Under favourable conditions, more than 90% of weeds are detected and not more than 25% of controlled weeds exhibit regrowth.

Recently, the image processing developed for the Rumex robot has been used to detect broad-leaved weeds on soccer fields. In this second project, an implement was built that can be towed by a quad or a small tractor. On the soccer field, detected weeds are controlled by application of the selective herbicide MCPA. Preliminary tests indicate a detection accuracy of 80%.