Search

This paper presents a novel machine vision-based approach detecting and mapping gall mite infection in dormant buds on black currant bushes. A vehicle was fitted with four cameras and RTK-GPS. Results compared automatic detection to human decisions based on the images, and by mapping the results into the same GPS map as the original ground truth recorded in the field. Human inspection found 52% of the infected bushes and the algorithm found the same 52%. The algorithm marked 7% of the healthy bushes, due to healthy buds looking larger closer to the camera. Consequently, the remaining challenge lies in optimising the viewing perspective.