Implementación de un sistema de clasificación de banano basado en el estado de maduración

Abstract

The Ecuadorian banana industry, as a pillar of the national economy, faces the challenge of modernizing its processes to meet the growing demands of the international market. This support project focuses on the implementation of an automated banana grading system based on artificial vision, capable of identifying the ripening stage of the fruit as they circulate on a conveyor belt. The proposed solution combines advanced image processing and deep learning techniques to optimize sorting accuracy and speed. Initially, it is necessary to highlight that an analysis of the state of the art in computer vision systems applied to banana classification was carried out, identifying methodologies, neural network architectures and preprocessing strategies relevant to the proposed problem. Subsequently, a representative dataset was developed, composed of high-resolution images of bananas in a controlled environment, maintaining uniform illumination conditions, fixed orientation and a single background. This dataset was used to train a convolutional neural network by means of transfer learning techniques, using pre-trained models in MobileNet, adjusted to optimize performance for the specific task. Finally, the performance of the system was evaluated in a controlled environment, using metrics such as accuracy, sensitivity, specificity and processing time, in order to measure the effectiveness of the model in classifying bananas at different stages of maturity. The results obtained seek to demonstrate that the implementation of artificial vision systems in the banana industry not only improves operational efficiency, but also positions the sector as a competitive and sustainable player in the global arena. This work lays the foundation for future applications of artificial intelligence technologies in the modernization of agricultural and agro-industrial processes.