A carregar...
Publicação
Monitorização do comportamento ergonómico de operadores na indústria 5.0 através de visão computacional baseada em Inteligência Artificial
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 23.96 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A presente dissertação propõe o desenvolvimento de um sistema não invasivo para monitorização do comportamento de operadores dentro do contexto da ergonomia, alinhado aos princípios da Indústria 5.0, que coloca o bem-estar humano no centro dos processos produtivos. O sistema visa identificar posturas inadequadas e prevenir Lesões Músculo-Esqueléticas Relacionadas com o Trabalho (LMERT), utilizando técnicas de Inteligência Artificial e Visão Computacional. Para tal, foi implementado um módulo de deteção de pose baseado no modelo MediaPipe Pose e uma câmara de profundidade Intel RealSense D415, que permite calcular ângulos posturais tridimensionais em tempo real. O software foi desenvolvido em Python, integrando módulos de filtragem temporal, persistência de eventos e uma base de dados SQLite, além de uma interface web interativa criada com Flask e WebSocket. O sistema foi validado experimentalmente por comparação com uma avaliação ergonómica conduzida por uma especialista, obtendo acurácia global de 94,6% e sensibilidade de 93,1%. Os resultados confirmam a eficácia da solução proposta na deteção automática de posturas inadequadas. O sistema contribui para o avanço da ergonomia digital e para a promoção da saúde ocupacional na Indústria 5.0, oferecendo uma ferramenta acessível, portátil e capaz de apoiar a prevenção de LMERT de forma contínua e em tempo real.
This dissertation proposes the development of a non-invasive system for monitoring operator behavior within the context of ergonomics, aligned with the principles of Industry 5.0, which places human well-being at the center of production processes. The system aims to identify inadequate postures and preventWork-Related Musculoskeletal Disorders (WMSDs), using Artificial Intelligence and Computer Vision techniques. To this end, a pose detection module based on the MediaPipe Pose model and an Intel RealSense D415 depth camera were implemented, which allows for the calculation of three-dimensional postural angles in real-time. The software was developed in Python, integrating modules for temporal filtering, event persistence, and an SQLite database, in addition to an interactive web interface created with Flask and WebSocket. The system was experimentally validated by comparison with an ergonomic assessment conducted by a specialist, achieving an overall accuracy of 94.6% and a sensitivity of 93.1%. The results confirm the effectiveness of the proposed solution in the automatic detection of inadequate postures. The system contributes to the advancement of digital ergonomics and the promotion of occupational health in Industry 5.0, offering an accessible, portable tool capable of supporting WMSD prevention continuously and in real-time.
This dissertation proposes the development of a non-invasive system for monitoring operator behavior within the context of ergonomics, aligned with the principles of Industry 5.0, which places human well-being at the center of production processes. The system aims to identify inadequate postures and preventWork-Related Musculoskeletal Disorders (WMSDs), using Artificial Intelligence and Computer Vision techniques. To this end, a pose detection module based on the MediaPipe Pose model and an Intel RealSense D415 depth camera were implemented, which allows for the calculation of three-dimensional postural angles in real-time. The software was developed in Python, integrating modules for temporal filtering, event persistence, and an SQLite database, in addition to an interactive web interface created with Flask and WebSocket. The system was experimentally validated by comparison with an ergonomic assessment conducted by a specialist, achieving an overall accuracy of 94.6% and a sensitivity of 93.1%. The results confirm the effectiveness of the proposed solution in the automatic detection of inadequate postures. The system contributes to the advancement of digital ergonomics and the promotion of occupational health in Industry 5.0, offering an accessible, portable tool capable of supporting WMSD prevention continuously and in real-time.
Descrição
Mestrado de dupla diplomação com a UTFPR - Universidade Tecnológica Federal do Paraná
Palavras-chave
Indústria 5.0 Ergonomia Inteligência artificial Visão computacional