[4-1600-D] Other Categories (1)

[4-1600-D-09] Onion Bulb Counting in a Large-scale Field Using a Drone with RTK-GNSS

*Satoshi Yamamoto¹, Hirokazu Madokoro¹, Yo Nishimura¹, Yukio Yaji¹ (1. Akita Prefectural University(Japan))

Keywords:Onion, Drone, RTK-GNSS, Orthomosaic, Deep learning

RTK-GNSS mounted on a drone made it possible that relatively accurate orthomosaic and digital elevation model (DEM) were generated with minimum number of ground control point (GCP) in a large-scale onion field. Deep learning was then applied to detect onion bulbs in the orthomosaic images. We annotated the bulbs of 16,812 with 250 images for the object detection. After machine learning process, the detection rate was 0.60 while the recall was 0.16. When the trained model was applied to the orthomosaic images, the number of the onion bulb ranged from 29,970 to 109,694 which were from 5 % to 19 % of the estimated onion plant number. Our next step would be to improve the trained model by reducing the false-negative value in the confusion matrix.