Publication: Algoritmo inteligente para el reconocimiento y acotación semántica de primitivas en una nube de puntos 3D
Loading...
Identifiers
Publication date
2015-07
Defense date
2016-10-21
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Describir detalladamente una escena digital, tal y como la percibe un humano no es una tarea fácil, se requiere de muchos factores como son conocimiento previo, capacidad de análisis y reconocimiento. Factores que no es posible encontrarlos en una sola ciencia, de esta manera, la hibridación del uso de maquetas digitales, la ingeniería inversa y tecnologías semánticas puede mejorar el proceso. El objetivo general de la tesis es definir un algoritmo de ingeniería inversa, capaz de reconocer elementos a partir de primitivas geométricas, específicamente cilindros, toros y esferas en una nube de puntos captados vía un escáner de escenas y asociarlos lógica, semántica y físicamente con elementos que se pueden encontrar en una escena real. Esto permitiría la generación de una maqueta digital enriquecida en donde los objetos serían descritos con un alto grado de detalle, proporcionando no solo la forma, sino también características geométricas como alto, largo, diámetro; su topología e información semántica de acuerdo al rol que representa cada elemento en la escena. En esta tesis se presenta en algoritmo propuesto, una implementación del mismo en un contexto industrial real y una evaluación que mida la calidad de la clasificación de dicho algoritmo comparándolo como un algoritmo de reconocimiento de formas geométricas tradicional.
Describing in detail a digital scene, as it is perceived by a human is not an easy task, it requires many factors such as prior knowledge, analytical skills and recognition. These factors cannot find a single science, in this sense, hybridization of using digital models, reverse engineering and semantic technologies can improve this process. The thesis overall aim is defining a reverse engineering algorithm, able to recognize elements from geometric primitives, specifically cylinders, tori and spheres in a cloud of points captured via a scanner, and associate them to logic, semantics and physic information to describe a real scene. This will allow the generation of a digital mockup-enriched where objects are described with a high degree of detail, providing not only the shape but also geometric characteristics such as height, length, diameter; its topology and semantic information according to the role that represents each item in the scene. In this thesis an intellgent algorithm, and its implementation in a real industrial context is presented. Also an assessment that measures the quality of the classification of this algorithm, comparing it with a traditional recognition algorithm is introduced.
Describing in detail a digital scene, as it is perceived by a human is not an easy task, it requires many factors such as prior knowledge, analytical skills and recognition. These factors cannot find a single science, in this sense, hybridization of using digital models, reverse engineering and semantic technologies can improve this process. The thesis overall aim is defining a reverse engineering algorithm, able to recognize elements from geometric primitives, specifically cylinders, tori and spheres in a cloud of points captured via a scanner, and associate them to logic, semantics and physic information to describe a real scene. This will allow the generation of a digital mockup-enriched where objects are described with a high degree of detail, providing not only the shape but also geometric characteristics such as height, length, diameter; its topology and semantic information according to the role that represents each item in the scene. In this thesis an intellgent algorithm, and its implementation in a real industrial context is presented. Also an assessment that measures the quality of the classification of this algorithm, comparing it with a traditional recognition algorithm is introduced.
Description
Keywords
Algoritmo inteligente, Reconocimiento, Acotación semántica, Nube de puntos