Computer vision-based system for quality management of Cavendish Banana in post-harvest stage

Abstract

Currently, the management of fruits in their post-harvest stage has been quite neglected by the Panamanian market; while large amounts are invested in planting and harvesting, no systems have been implemented to allow better management of these. Therefore, resulting in large economic losses combined with unused products that are discarded daily in sales centers by traders of various levels. With this in mind, we seek to develop a system based on computer vision to determine the ripening stage of a fruit and estimate its shelf life. This study focuses on the banana as a highly demanded fruit with several maturation stages. During the development, the temperature and humidity will be captured, and recording the elapsed time from its green stage to its rotting point, to generate a machine learning model that, given an image of an input banana, will infer as an answer the approximate time for reaching a point where it cannot be consumed or used to make derivative products. Once the models were created, these obtained satisfactory results that were checked with the reality of the banana's lifespan. The tests were carried out using a mobile application that allowed obtaining the estimates in real-time. Both, the application and the developed models, allowed to adequately estimate the ripening stage and shelf life of the fruit using artificial vision and machine learning algorithms.