Desarrollo de un sistema de visión artificial basado en redes neuronales convolucionales para la detección del estado de maduración del cacao mediante una aplicación móvil

Abstract

This project describes the design and implementation of an intelligent system based on computer vision to determine the ripeness stage of cocoa pods through a mobile application. Given the importance of fine aroma cocoa in Ecuador and the need to optimize harvesting processes, a technological solution is proposed to automate the identification of pod maturity stages directly in the cultivation environment. For this purpose, real images captured in plantations were classified and labeled to train an image processing model. The model training was conducted using a supervised approach, distinguishing between immature, ripening, and mature pods, achieving high levels of accuracy after several iterations (96 % on average). Subsequently, the model was exported and optimized for integration into an interactive web application, allowing users to upload, process them locally, and obtain the visual classification of each pod’s maturity stage. As a result, a functional, accessible, and adaptable system was developed, aimed at small and medium-sized producers, to facilitate decision-making regarding the optimal harvesting time, reduce losses due to premature or late collection; and promote precision agriculture in rural areas. The proposed methodology, in addition to its applicability to cocoa, can be adapted to other crops requiring visual fruit analysis, contributing to the digitalization of the agricultural sector and fostering sustainable and efficient practices within production chains.