• 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.55.2023.tde-05092023-090640
Document
Author
Full name
Jean Amaro
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2023
Supervisor
Committee
Osório, Fernando Santos (President)
Grassi Junior, Valdir
Pessin, Gustavo
Todt, Eduardo
Title in Portuguese
3D-CSD+: Extração de características 3D baseada em grafos
Keywords in Portuguese
Extração de características 3D
Reconhecimento de padrões
Visão computacional
Abstract in Portuguese
A área da Robótica é uma das que se beneficia do desenvolvimento tecnológico recente, testemunhando um crescente interesse no desenvolvimento de novas aplicações em diferentes áreas. Em muitas delas, a Visão Computacional desempenha um papel importante, uma vez que muitos robôs dependem de câmeras para o seu funcionamento. Com o desenvolvimento tecnológico, estão hoje disponíveis sensores capazes de obter dados tridimensionais, motivando o desenvolvimento de algoritmos de percepção nesse plano dimensional. Este trabalho de mestrado propõe uma nova técnica para a descrição de objetos 3D, de maneira factível com aplicações limitadas computacionalmente. As características (features) extraídas são robustas e invariantes a transformações (p.ex. translação, rotação e mudança de escala), e que permite o reconhecimento de objetos em aplicações embarcadas e com requisitos de tempo real. Em testes usando o dataset ModelNet40, chegou-se a uma taxa de acerto Top-3 de 80%, com menos de 20ms de execução por amostra.
Title in English
3D-CSD+: Graph based 3D feature extraction
Keywords in English
3D Feature extraction
Computer vision
Pattern recognition
Abstract in English
The field of Robotics is one of those that benefits from recent technological development, witnessing a growing interest in the development of new applications in various areas. In many of these areas, Computer Vision plays an important role, as many robots rely on cameras for their operation. With technological advancement, sensors capable of obtaining three-dimensional data are now available, motivating the development of perception algorithms in this dimensional plane. This masters thesis proposes a new technique for the description of 3D objects, feasible for computationally limited applications. The extracted features are robust and invariant to transformations (e.g., translation, rotation, and scale changes), enabling object recognition in embedded applications with real-time requirements. In tests using the ModelNet40 dataset, a Top-3 accuracy rate of 80% was achieved, with less than 20ms of execution time per sample.
 
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
2023-09-05
 
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.