Weed ID, Mapping and Spraying System
In broadacre farming, weeds and crops compete for sunlight, moisture and nutrients in order to grow.
Therefore, to maximise crop growth, herbicide is applied to control weeds using conventional spraying systems. These systems are incapable of identifying weeds and must apply herbicide across the entire paddock to ensure all weeds are controlled. To effectively control weeds without wasting herbicide, a targeted Weed Identification, Mapping and Spraying System (WIMSS) has been developed.
WIMSS implements computer vision and machine learning to identify, map and selectively spray weeds in real-time. A camera mounted on a conventional spray boom scans the crop ahead, weeds are detected using a trained deep convolutional neural network (DCNN), weed locations are mapped using RTK GPS coordinates, and individual spray nozzles are controlled to achieve targeted spraying. WIMSS is able to identify weeds with 85% accuracy, map weed locations with 2cm accuracy, and spray at a speed of 5km/hr using 60% less herbicide than conventional methods.