| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 26.04 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A qualidade desempenha um papel fundamental na indústria moderna, impactando
diretamente a eficiência, a competitividade entre empresas e a satisfação do
cliente. Tradicionalmente, a inspeção de qualidade é realizada por operadores humanos,
estando por isso sujeita a limitações como fadiga, subjetividade e inconsistências.
Com os avanços tecnológicos, surgiram sistemas automáticos de inspeção
visual que utilizam visão computacional para minimizar esses erros humanos, proporcionando
inspeções mais rápidas, precisas e repetíveis.
Nesse contexto, a Schmitt-Elevadores propôs o desenvolvimento de um sistema de
inspeção visual automático para detetar defeitos em rampas móveis, um componente
que integra o sistema de abertura de portas do elevador. Esse sistema visa verificar
a montagem e o correto funcionamento mecânico das mesmas por meio da análise de
imagens e vídeos. Atualmente, na empresa, a inspeção desse componente é realizada
manualmente, dependendo da experiência do operador e sujeita a variações, sendo
um dos principais objetivos a redução da subjetividade do processo para garantir
uma qualidade ainda melhor.
A solução final é uma aplicação web que combina técnicas de visão computacional
tradicional e deteção de objetos por inteligência artificial. O processo pode ser
dividido em duas fases: a primeira, de inspeção estática, captura uma imagem da
peça e a compara-a com uma imagem de referência para avaliar sua conformidade
visual; a segunda, de análise dinâmica, grava um vídeo do movimento da peça e
verifica se o comportamento está dentro dos padrões esperados.
A solução foi validada por meio de testes experimentais, cujos resultados mostraram
que o sistema é capaz de detetar defeitos e avaliar corretamente o movimento,
atendendo aos objetivos propostos. Durante o desenvolvimento foram identificadas
limitações relacionadas à iluminação, ao posicionamento da peça e à sensibilidade
do sistema a pequenas variações, o que destaca a importância da padronização do
ambiente de captura.
Quality plays a big role in modern industry, directly impacting efficiency, competitiveness between companies, and customer satisfaction. Traditionally, quality inspection has been carried out by human operators, which is subject to limitations such as fatigue, subjectivity, and inconsistencies. With technological advancements, automated visual inspection systems that use computer vision have emerged to minimize human errors, providing faster, more accurate, and repeatable inspections. In this context, Schmitt-Elevadores proposed the development of an automatic visual inspection system to detect defects in rampas móveis, a component that integrates the elevator’s door opening system. This system aims to verify the assembly and proper mechanical functioning of the rampas through the analysis of images and videos. Currently, the company performs this inspection manually, relying on the operator’s experience and subject to variations, one of the main objectives is the reduction of subjectivity to ensure an even higher quality. The final solution is a web application that combines traditional computer vision techniques and object detection using artificial intelligence. The process can be divided into two phases: the first, a static inspection, captures an image of the part and compares it with a reference image to assess its visual conformity; the second, dynamic analysis, records a video of the part’s movement and checks whether the behavior aligns with expected standards. The solution was validated through experimental tests, and the results showed that the system is capable of detecting defects and correctly evaluating movement, meeting the proposed objectives. During the development, limitations related to lighting, positioning, and the system’s sensitivity to small variations were identified, highlighting the importance of standardizing the capture environment.
Quality plays a big role in modern industry, directly impacting efficiency, competitiveness between companies, and customer satisfaction. Traditionally, quality inspection has been carried out by human operators, which is subject to limitations such as fatigue, subjectivity, and inconsistencies. With technological advancements, automated visual inspection systems that use computer vision have emerged to minimize human errors, providing faster, more accurate, and repeatable inspections. In this context, Schmitt-Elevadores proposed the development of an automatic visual inspection system to detect defects in rampas móveis, a component that integrates the elevator’s door opening system. This system aims to verify the assembly and proper mechanical functioning of the rampas through the analysis of images and videos. Currently, the company performs this inspection manually, relying on the operator’s experience and subject to variations, one of the main objectives is the reduction of subjectivity to ensure an even higher quality. The final solution is a web application that combines traditional computer vision techniques and object detection using artificial intelligence. The process can be divided into two phases: the first, a static inspection, captures an image of the part and compares it with a reference image to assess its visual conformity; the second, dynamic analysis, records a video of the part’s movement and checks whether the behavior aligns with expected standards. The solution was validated through experimental tests, and the results showed that the system is capable of detecting defects and correctly evaluating movement, meeting the proposed objectives. During the development, limitations related to lighting, positioning, and the system’s sensitivity to small variations were identified, highlighting the importance of standardizing the capture environment.
Description
Keywords
Automatic visual inspection Quality control Computer vision YOLO OpenCV Bounding box Inspeção visual automática Controlo de qualidade Visão computacional
Pedagogical Context
Citation
Publisher
CC License
Without CC licence
