Por favor, use este identificador para citar o enlazar este ítem: http://cio.repositorioinstitucional.mx/jspui/handle/1002/1322
RGB AND MULTISPECTRAL IMAGE ANALYSIS BASED ON DEEP LEARNING FOR REAL-TIME DETECTION AND CONTROL OF WEEDS IN CORNFIELDS
Francisco Garibaldi Márquez
Luis Manuel Valentín Coronado
Gerardo Flores
Acceso Abierto
Atribución-NoComercial-SinDerivadas
Deep learning
Weed detection
Weed control
Smart weed sprayer
Reduction of herbicide usage
"In this study, a vision system based on deep learning (DL) was proposed for real-time detection and control of weeds in actual corn fields. Initially, a dataset comprising RGB and multispectral images was generated and annotated at the pixel level. Subsequently, both end-to-end semantic segmentation convolutional neural networks (CNNs) and transformers were investigated. The transformer model outperformed CNNs in segmenting weeds. Utilizing this vision system, a Smart Weed Sprayer (SWS) was developed, resulting in a 45.64% reduction in herbicide usage compared to a conventional weed sprayer (CWS), while maintaining similar effectiveness in weed control."
23-04-2024
Tesis de doctorado
Inglés
Bibliotecarios
Estudiantes
Investigadores
Público en general
Garibaldi-Márquez, (2024). "RGB and multispectral image analysis based on deep learning for real-time detection and control of weeds in cornfields". Tesis de Doctorado Interinstitucional en Ciencia y Tecnología. Centro de Investigaciones en Óptica, A.C. Aguascalientes, Ags., México. 214 páginas.
PROTECCIÓN DE LOS CULTIVOS
Versión publicada
publishedVersion - Versión publicada
Aparece en las colecciones: DOCTORADO INTERINSTITUCIONAL EN CIENCIA Y TECNOLOGÍA (DPICYT)

Cargar archivos:


Fichero Descripción Tamaño Formato  
18467.pdfTexto completo/ Full-text available3.95 MBAdobe PDFVisualizar/Abrir