Online weed control using digital image analysis

This paper presents a real-time approach for site-specific weed control in winter barley, sugar beet and maize fields. A digital image analysis system was used to assess weed seedling distribution and plant species composition in the field. The system consists of three CCD-cameras mounted on a spray boom in front of a vehicle. The images of all three cameras were analyzed automatically. Weed and crop species were identified using shape analysis. Weed species were grouped according to their competitive ability. A simple decision rule was used to turn on each section of the spray boom separately when a threshold for weed infestation was exceeded. A differential GPS receiver was installed on the vehicle to document the application maps. Site-specific weed control resulted in an average herbicide reduction in winter barley for broad-leaved weeds of 60 % and for grass-weeds of 92 %, in maize for broad-leaved weeds 11% and for grass-weeds 81 % and in sugar beet for broad-leaved weeds 42 % and for grass-weeds 36%.