Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 28, 2023 - July 01, 2023
This paper presents a method of flower detection by deep learning in automatic pollination system using drone. Most of plants are pollinated by bees. However, the number of bees decrease because of global warning etc., moreover, non-native bees have an ecosystem impact. For these reasons, some pollination robots have been developed. We aim to develop a pollination system using drone that has small body and high mobility as alternative bees. Drone control, flower detection, and a pollination equipment are the element technologies for developing this system. This paper compares four flower detection models depending on the network size using YOLO (You Only Look Once) v5. We evaluate the result in terms of detection accuracy and speed and show the effective model for the pollination system using drone.
