Segmentación en imagen de frutos de granado usando deep learning con aplicación en agricultura de precisión
- Jaime Giménez Gallego 1
- Juan Domingo González Teruel 1
- Ana Toledo Moreo 1
- Manuel Jiménez Buendía 1
- Fulgencio Soto Vallés 1
- Roque Torres Sánchez 1
-
1
Universidad Politécnica de Cartagena
info
- Carlos Balaguer Bernaldo de Quirós (ed. lit.)
- José Manuel Andújar Márquez (ed. lit.)
- Ramón Costa Castelló (ed. lit.)
- C. Ocampo-Martínez (coord.)
- Juan Jesús Fernández Lozano (ed. lit.)
- Matilde Santos Peñas (ed. lit.)
- José Simo (ed. lit.)
- Montserrat Gil Martínez (ed. lit.)
- José Luis Calvo Rolle (ed. lit.)
- Raúl Marín (ed. lit.)
- Eduardo Rocón de Lima (ed. lit.)
- Elisabet Estévez Estévez (ed. lit.)
- Pedro Jesús Cabrera Santana (ed. lit.)
- David Muñoz de la Peña Sequedo (ed. lit.)
- José Luis Guzmán Sánchez (ed. lit.)
- José Luis Pitarch Pérez (ed. lit.)
- Óscar Reinoso García (ed. lit.)
- Óscar Déniz Suárez (ed. lit.)
- Emilio Jiménez Macías (ed. lit.)
- Vanesa Loureiro-Vázquez (ed. lit.)
Publisher: Servizo de Publicacións ; Universidade da Coruña
Year of publication: 2022
Pages: 1001-1006
Congress: Jornadas de Automática (43. 2022. Logroño)
Type: Conference paper
Abstract
In precision agriculture, to automatically monitor the state of the crop using images, processing tools are needed to extract the information of interest. In this study, a Deep Learning model is developed for image segmentation to discriminate pomegranate fruits. Results of Intersection over Union (IoU)=0.71 and mean Average Precision (mAP)=0.82 are achieved. Subsequently, an algorithm for estimating the size of the fruit in pixels is presented, with an average relative error of 5.4%.