/img alt="Imagem da capa" class="recordcover" src="""/>
Tese
Identificação e quantificação de desplacamento cerâmico em fachadas de edifícios no contexto da indústria 4.0
The facade maintenance process is guided by the results obtained in the inspection phase. Some proposals for methods aimed at improving the inspection process have been discussed, and among these, those that are conducted based on Digital Image Processing (PDI) techniques captured by Unmanned Aer...
Autor principal: | SOUSA, Alcineide Dutra Pessoa de |
---|---|
Grau: | Tese |
Idioma: | por |
Publicado em: |
Universidade Federal do Pará
2023
|
Assuntos: | |
Acesso em linha: |
https://repositorio.ufpa.br/jspui/handle/2011/15773 |
Resumo: |
---|
The facade maintenance process is guided by the results obtained in the inspection phase.
Some proposals for methods aimed at improving the inspection process have been
discussed, and among these, those that are conducted based on Digital Image Processing
(PDI) techniques captured by Unmanned Aerial Vehicle (UAV) stand out. The use of
UAVs to capture images on facades streamlines access to the inspected area, and PDI
techniques help to automate the process of identifying pathological manifestations. In
addition, the fourth industrial revolution has allowed the use of various technological tools
in the most varied engineering applications. Among these technologies we can mention
cloud computing and computer vision algorithms. In this context, this research aims to
apply PDI techniques to detect regions with ceramic displacement on building facades
using technologies relevant to industry 4.0 (fourth industrial revolution). The
methodological procedure used starts with the formation of a database (images) captured
by cell phone and UAV. For modeling purposes, the YOLO (You Only Look Once) object
detection algorithm was applied to the images that make up the database using cloud
computing. The applied methodology resulted in a program written in python capable of
identifying the regions with displacement, quantifying the missing ceramics and exporting
the quantification results in a spreadsheet. The identification process had success rates
close to 99% and the quantification errors of less than one ceramic per image, which leads
to the conclusion of the feasibility of the proposed computational program |