Precision weed control robot for vegetable fields with high crop and weed densities

Abstract

Abstract. Weeds growing in close proximity to the seedline area are highly competitive for resources needed by crop plants and causes a reduction in crop productivity if left uncontrolled. A method to automate the weed removal process, which is currently a costly manual operation, is in high demand by vegetable growers. Automatic weed removal requires machine vision, displacement sensing, and actuation systems. For high plant and weed density scenarios common in organic production of leafy greens, the development of a machine vision system that analyzes images to provide weed and crop mapping while spraying actuators are required to automatically spray on the weed location precisely depending on the displacement. It is a challenging task to develop a high-resolution spraying actuator and automatically controls it while synchronizing it with all other subsystems to work in the actual field at high-speed. This paper describes a spraying actuator that was developed to be controlled in real-time based on weed and crop mapping. To allow for image processing time, and to minimize spray drift collecting on the camera lens, the spray nozzles are placed 0.8 m behind the camera. The performance of spraying actuator is tested and evaluated in a densely populated lettuce field scenario with a selective herbicide dosing treatment area of 1 x 1 cm at the continuous speed of 2.4 km/h. The spraying actuator is able to spray correctly detected weeds without spray drift. The results established that the developed herbicide spraying actuator was robust and effective for the automatic weed control process.