Diseño de sistema de inspección automatizada de calidad de granos de cacao mediante visión artificial

Abstract

CONDITION FOR PUBLICATION OF PROJECT. Fine-flavor cocoa is one of Ecuador’s main export products; however, manual bean classification still used in post-harvest processes presents limitations associated with subjectivity, operator fatigue, and low efficiency. In this context, the project aims to design and build a prototype of an automated inspection system, based on computer vision and a YOLO-based artificial intelligence model, to detect physical defects in cocoa beans and optimize the selection process. A laboratory prototype was developed, consisting of a conveyor belt, a controlled lighting system, an area-scan camera, the classification model, and a pneumatic ejection mechanism, all integrated through an ESP32 microcontroller. Tests demonstrated stable performance of the subsystems and a classification accuracy above 90 % under controlled conditions. It is concluded that the proposed system constitutes a viable and replicable alternative to improve efficiency and reliability in cocoa bean selection processes. Keywords: computer vision, automated classification, cocoa, artificial intelligence.