| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 25.05 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A análise de circuitos elétricos/eletrónicos é uma componente fundamental na área da Engenharia Eletrotécnica. Geralmente, os estudantes iniciam a análise de um circuito desenhando-o em papel, e, posteriormente à sua resolução (analítica), implementam o circuito num simulador para verificar e validar os resultados obtidos analiticamente. No entanto, este processo faz com que os estudantes necessitem de voltar a desenhar o circuito, tarefa que consome tempo e durante a qual poderão ser introduzidos erros ao traçar o mesmo no simulador.
Este projeto consiste no desenvolvimento de um módulo de software capaz de interpretar a topologia de um esquema elétrico e de construir o seu modelo a partir de uma imagem, quer seja desenhado “à mão” ou a computador. Este módulo é baseado em algoritmos de visão computacional, propondo uma abordagem inovadora para a análise geral de um circuito elétrico através de uma imagem, com a deteção e extração dos seguintes elementos presentes no mesmo: componentes, nós, interligações entre os elementos, e labels (nomes, valores, etc) associadas aos respetivos elementos do circuito. O software deste projeto gera um mapa com o modelo do esquema elétrico analisado, resultante do processo de segmentação efetuado, bem como imagens com a region of interest (ROI) de cada componente e label extraída. Os ficheiros produzidos poderão ser utilizados por um outro módulo de software (não abrangido pela presente dissertação), responsável pela classificação dos componentes segmentados e também pelo reconhecimento dos caracteres alfanuméricos contidos nas labels do circuito. Assim, a combinação de ambos os módulos de software conceberá uma aplicação completa de reconhecimento de circuitos elétricos a partir de uma imagem. A finalidade futura desta aplicação é a sua integração na ferramenta U=RIsolve, framework web para ensino e auto-aprendizagem de metodologias para análise de circuitos, o que permitirá aos estudantes a obtenção e validação dos resultados do circuito através de uma imagem.
The analysis of electrical/electronic circuits is a fundamental component in the area of Electrical Engineering. Generally, students start the analysis of a circuit by drawing it on paper, and, after its (analytical) resolution, they implement the circuit in a simulator to verify and validate the results obtained analytically. However, this process makes students need to redraw the circuit, a time-consuming task during which errors may be introduced when tracing the circuit in the simulator. This project consists in the development of a software module capable of interpreting the topology of an electrical schematic and of building its model from an image, whether drawn by hand or by computer. This module is based on computer vision algorithms, proposing an innovative approach for the general analysis of an electrical circuit through an image, with the detection and extraction of the following elements present in it: components, nodes, interconnections between elements, and labels (names, values, etc) associated with the respective circuit elements. The software of this project generates a map with the model of the electrical schematic analyzed, resulting from the segmentation process carried out, as well as images with the region of interest (ROI) of each component and label extracted. The files produced may be used by another software module (not covered by this dissertation), responsible for classifying the segmented components and also for recognizing the alphanumeric characters contained in the labels of the circuit. Thus, the combination of both software modules will create a complete application for recognizing electrical circuits from an image. The future purpose of this application is its integration into the U=RIsolve tool, a web framework for teaching and self-learning methodologies for circuit analysis, which will allow students to obtain and validate the circuit results through an image.
The analysis of electrical/electronic circuits is a fundamental component in the area of Electrical Engineering. Generally, students start the analysis of a circuit by drawing it on paper, and, after its (analytical) resolution, they implement the circuit in a simulator to verify and validate the results obtained analytically. However, this process makes students need to redraw the circuit, a time-consuming task during which errors may be introduced when tracing the circuit in the simulator. This project consists in the development of a software module capable of interpreting the topology of an electrical schematic and of building its model from an image, whether drawn by hand or by computer. This module is based on computer vision algorithms, proposing an innovative approach for the general analysis of an electrical circuit through an image, with the detection and extraction of the following elements present in it: components, nodes, interconnections between elements, and labels (names, values, etc) associated with the respective circuit elements. The software of this project generates a map with the model of the electrical schematic analyzed, resulting from the segmentation process carried out, as well as images with the region of interest (ROI) of each component and label extracted. The files produced may be used by another software module (not covered by this dissertation), responsible for classifying the segmented components and also for recognizing the alphanumeric characters contained in the labels of the circuit. Thus, the combination of both software modules will create a complete application for recognizing electrical circuits from an image. The future purpose of this application is its integration into the U=RIsolve tool, a web framework for teaching and self-learning methodologies for circuit analysis, which will allow students to obtain and validate the circuit results through an image.
Description
Keywords
Análise de circuitos Visão computacional Processamento de imagem Segmentação de circuito Aprendizagem Circuit analysis Computer vision Image processing Circuit Segmentation Learning