• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.95.2021.tde-28052021-173908
Document
Author
Full name
Juliana Virginio da Silva
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2021
Supervisor
Committee
Costa, Luciano da Fontoura (President)
Borges, Leonardo Maurici
Marana, Aparecido Nilceu
Travieso, Gonzalo
Title in Portuguese
Análise e reconhecimento de padrões em feixes vasculares vegetais
Keywords in Portuguese
Atributos
Feixes vasculares
Histologia vegetal
Visão computacional
Abstract in Portuguese
Os tecidos vasculares ou tecidos condutores são os responsáveis pelo transporte de água, sais minerais e nutrientes ao longo do vegetal. A literatura evidencia a existência de cinco tipos de feixes vasculares vegetais sendo eles: colateral aberto, colateral fechado, bicolateral, anficrival e anfivasal. Tradicionalmente, a identificação desses tipos de feixes é feita por especialistas em anatomia vegetal. Contudo, os avanços na área de análise de imagens digitais vem abrindo caminho para que estudo morfológico vegetal ganhem novas ferramentas de análise. Os avanços na ciência também têm promovido a interdisciplinaridade em pesquisas e isso tem gerado novas aplicações e abordagens. Partindo dessa premissa, este trabalho analisou imagens microscópicas dos diversos feixes vasculares com o objetivo de promover um estudo capaz de 1) prover subsídios para segmentação em imagens microscópicas de vegetais e 2) contribuir para trabalhos de identificação tecidual. A aquisição das imagens foi feita por meio de microscopia óptica. As lâminas histológicas preparadas com cortes transversais de caules frescos, de espécies representativas de cada tipo de feixe vascular foram visualizadas e digitalizadas. As imagens foram segmentadas manualmente e foram extraídos atributos de quatro categorias: Forma, densidade, regularidade estrutural e multiescala. As matrizes contendo os valores obtidos foram submetidas a análise individual, análise par a par e análise de componentes principais, com o objetivo de selecionar atributos relevantes para a caracterização de tecidos vasculares. Todas as categorias de atributos foram avaliadas pelas abordagens k-Nearest Neighbors e Perceptron Multicamada, tanto em suas composições originais quanto após aplicação da análise de componente principal. Ao final, um conjunto de atributos foi selecionado. Esse conjunto obteve 91,67% de acertos em ambos classificadores ou seja, foi o que obteve melhor desempenho classificatório dos feixes vasculares vegetais. Um outro resultado relevante foi que o uso de atributos multiescala levou a uma porcentagem de acerto maior (81,66%) do que atributos tradicionalmente usados para descrever feixes vasculares, como por exemplo descritores de forma (70,84%). Ademais, nosso trabalho também estudou as possíveis explicações biológicas dos descritores selecionados. Isso indica que o presente trabalho foi capaz, não só, de elencar atributos capazes de classificar com bom desempenho os feixes vasculares vegetais, como também apresentar interpretações biológicas que potencialmente justifiquem os resultados aqui apresentados.
Title in English
Analysis and pattern recognition in vegetable vascular bundles
Keywords in English
Attributes
Computer vision
Plant histology
Vascular bundles
Abstract in English
Vascular tissues or conducting tissues are responsible for transporting water, mineral salts and nutrients throughout the plant. The literature shows the existence of five types of vegetable vascular bundles, namely: open collateral, closed collateral, bicolateral, amphicrival and amphivasal. The science that studies plants tissues is known as plant histology. Traditionally, the identification of these types of bundles is done by specialists in plant anatomy. However, advances in the area of digital image analysis are paving the way for plant morphological studies to gain new analysis tools. Advances in science have also promoted interdisciplinarity in research and this has generated new applications and approaches. Based on this premise, this work analyzed microscopic images of the various vascular bundles in order to promote a study capable of 1) providing subsidies for segmentation in microscopic images of plants and 2) contributing to the work of tissue identification. The acquisition of the images was made by means of optical microscopy. We visualized and scanned the histological slides prepared with cross sections of fresh stems, of species representative of each type of vascular bundle. The images were manually segmented and attributes from four categories were extracted: Shape, density, structural regularity and multiscale. The matrices containing the attribute values were subjected to individual analysis, peer analysis and principal component analysis in order to select the attributes that were most relevant throughout the work. The four categories of attributes, the selected attributes and a set containing all the extracted descriptors were evaluated by the k-Nearest Neighbors and Multilayer Perceptron approaches, both in their original dimensions and after application of the main component analysis. In the end, the category that contained the selected attributes obtained the best classificatory performance of the vascular bundles, reaching 91.67% of correct answers in both classifiers. Another relevant result was that the multiscale category obtained a higher percentage of correct answers (81.66%) than categories traditionally used to describe vascular bundles, such as shape (70.84%). In addition, our work also studied the possible biological explanations of the selected descriptors. This indicates that the present work was able, not only, to list attributes capable of classifying plant vascular bundles with good performance, but also to present biological interpretations that potentially justify the results presented here.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2021-06-01
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.