Análisis morfológico de semillas con redes memristivas

Seed morphology analysis is an essential task for studying and classifying seeds, and for any ulterior phytosanitary application such as weed detection. In this work, the analysis of the morphology of seeds is done by means of a memristive grid which is used as an analog image processor in order to...

Descripción completa

Detalles Bibliográficos
Autor: Marco Antonio Zamudio Hernández
Tipo de recurso: tesis de maestría
Estado:Versión aceptada para publicación
Fecha de publicación:2020
País:México
Institución:Instituto Nacional de Astrofísica, Óptica y Electrónica
Repositorio:Repositorio Institucional del INAOE
Idioma:español
OAI Identifier:oai:inaoe.repositorioinstitucional.mx:1009/1972
Acceso en línea:http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/1972
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/Inspec/Memristor
info:eu-repo/classification/Inspec/Memristive grids
info:eu-repo/classification/Inspec/Edge detection
info:eu-repo/classification/Inspec/Parallel processing
info:eu-repo/classification/Inspec/Seed morphology
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/22
info:eu-repo/classification/cti/2203
Descripción
Sumario:Seed morphology analysis is an essential task for studying and classifying seeds, and for any ulterior phytosanitary application such as weed detection. In this work, the analysis of the morphology of seeds is done by means of a memristive grid which is used as an analog image processor in order to achieve edge detection on images from a seed database. The memristors within the grid are modeled by a memristance expression that is given as a charge-controlled function. Besides, the grid is implemented in a parallelized version with a GPU scheme. The solution of the grid is obtained by a time-domain analysis of the resulting differential equations from the nodal analysis. In a first step, with the aim of highlighting the advantages of such the parallel implementation with respect to its serial counterpart, a set of benchmark images has been treated for edge detection. Not only are the qualitative aspects of the results kept, but also the main quantitative merit figures are almost identical. On the top of this, the parallel memristive grid exhibits a sensible gain in execution times in comparison with the serial grid. The main goal of this work, i.e. the analysis of the morphology of seeds is achieved by resorting to a series of measures on the seed edges. A geometric model of the seed is introduced in the form of an ellipse that is valid for monocots as well as dicots, and its main dimensions are used for establishing a group of merit figures that define the morphology. Finally, several case studies are presented.