Coffee classification according to its detachment force: a decision tree-based approach

• Ano de defesa: 2023
• Autor(a) principal: Meneses, Mariana Dias. [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Machine Learning; Selective Mechanized Harvesting; Uneven Maturation; Coffee; Agricultura
• Link de acesso: http://hdl.handle.net/11449/242933

The world coffee consumption demands high-efficiency crop systems. Consumers appreciate flavor and aroma in this beverage, characteristics that are game-changing in coffee value. A key role to improve this production chain, mechanized harvesting fails in provide coffee fruits selectivity. It means that the industry receives fruits with astringent flavor or fermentation. Because coffee plant has uneven maturation, i.e., green, cherry, and dry fruits, and the harvester settings are generalist, the fruits are detached regardless their maturation stage. The use of Machine Learning techniques improves the traditional agriculture to a digital one, its use in mechanized harvesting enhances selectivity of the coffee fruits. Overall, the present study aimed to classify the coffee fruit detachment force using a Decision Tree Classifier. The experiment was conducted in two field in the Brazilian state of Minas Gerais. A dynamometer was used to measure the detachment force of 23 coffee cultivars. The cultivars were grouped using a cluster algorithm and a Decision Tree classified each group according to the detachment force. The Decision Tree obtained a mean Matthews Correlation Coefficient of 0.81, proving its efficiency in classify the detachment force. Therefore, we proved that Decision Tree can power the mechanized harvesting as a tool to more accurate decision-making settings