Article - Main Track

Dronapple: A model for apple picking using machine learning and drones

Author: BARBOSA, M. F; HELFER, G. A.; BARBOSA, J. L. V.

Abstract: This article presents the Dronapple model, which uses a drone equipped with a camera and a suction cup for apple orchards in the Tall Spindle system with the ability to recognize apples in trees, harvest them with a suction cup and deposit them in a pantry box, using algorithms of computer vision and reinforcement learning. The validation and training of the proposed model were carried out with the implementation of two simulation environments. The simulator results showed that it took an average of 5.32 seconds to harvest the apple and an average of 17.75 seconds to move the drone to the pantry box and return to the tree with a 97.93% success rate in harvesting ripe apples in unobstructed positions. A video¹ showing the model working in the developed simulator environment and the source code² are available.

Keywords: Apple orchard, Autonomous drone, Computer Vision, Reinforcement learning, Simulator.

Full paper (in Portuguese)

Full Reference: Barbosa, M. F; Helfer, G. A.; Barbosa, J. L. V., "Dronapple: Um Modelo para Colheita de Maçãs Utilizando Aprendizado de Máquina e Drones", Revista de Sistemas de Informação da FSMA n 31(2023) pp. 48-68

Back