万能吸着グリッパを用いたディープラーニングによるトマト自動収穫ロボットの開発

抄録

In recent years, the number and average age of agricultural workers have increased; the development of new methods in agriculture, utilizing robot technology and ICT, is expected. This study proposes an automatic tomatoharvesting system that combines fruit-image recognition and deep learning, which has been remarkably developed in recent years; a harvesting device that uses the universal vacuum gripper has also been proposed. To validate the proposed system, harvesting experiments were conducted on a single fruit and on a collection of fruits. In the former case, robotic harvesting achieved a success rate of 93%. The results validate the feasibility of the fruit-position-estimation system.