Bibliographic citations
Fiestas, E., (2022). Desarrollo de una estrategia inteligente mediante redes neuronales convolucionales profundas para la medición de la calidad del crecimiento de plantines producidos en viveros industriales de la región La Libertad [Tesis, Universidad Privada Antenor Orrego]. https://hdl.handle.net/20.500.12759/8841
Fiestas, E., Desarrollo de una estrategia inteligente mediante redes neuronales convolucionales profundas para la medición de la calidad del crecimiento de plantines producidos en viveros industriales de la región La Libertad [Tesis]. PE: Universidad Privada Antenor Orrego; 2022. https://hdl.handle.net/20.500.12759/8841
@mastersthesis{sunedu/3586316,
title = "Desarrollo de una estrategia inteligente mediante redes neuronales convolucionales profundas para la medición de la calidad del crecimiento de plantines producidos en viveros industriales de la región La Libertad",
author = "Fiestas Sorogastua, Erick M.",
publisher = "Universidad Privada Antenor Orrego",
year = "2022"
}
Agriculture in Peru is growing at an accelerated rate; this has allowed its industrialization. Industrial nurseries must provide seedlings to farmers, a year they must produce more than 400 million seedlings only artichoke. However, the increase in production is becoming more difficult due to the scarcity of specialized labor for quality control of products, within which there is a specific visual analysis procedure to classify seedlings according to their quality. The present thesis work presents the development of an intelligent strategy using deep convolutional neural networks for measuring the quality of the growth of seedlings produced in industrial nurseries in the region of La Libertad. First, the necessary dataset based on images of artichoke seedlings is prepared, including images of real and synthetic seedlings. Second, a set of Machine and Deep Learning models (PCA, kmeans, VGG16 and Yolov3) are proposed and designed in order to be able to perform a comparison of the performances of each of these for the task of detecting and classifying RGB images of the seedlings. Third, the training, testing and validation of the models is carried out. Fourth, the Pierson correlation is obtained to verify the relationship between the prediction of previously trained models and the classification criteria of an industrial nursery in the region. Finally, the model is displayed with a respective monitoring panel (dashboard) that runs on the local network and in the cloud (Google Cloud Platform). The conclusions of the work are shown as a Computer Vision System (SVC) based on Industrial Internet of Thing (IIoT) that integrates AI, industrial components, cloud computing and robotics, achieving a seedling classification capacity that correlates an 85 % with what has been done in industrial nurseries.
This item is licensed under a Creative Commons License