Sistema de monitoramento para determinação de desempenho em uma linha seriada com base na detecção de contornos e vértices com OpenCV

• Autor(a) principal: Bertelli, Pedro
• Data de Publicação: 2021
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UTFPR (da Universidade Tecnológica Federal do Paraná (RIUT))
• Texto Completo: http://repositorio.utfpr.edu.br/jspui/handle/1/27862
• topic: Sistemas embarcados (Computadores); Visão por computador; Automação; Embedded computer systems; Computer vision; Automation; CNPQ::ENGENHARIAS

The frequent evolutions in the market require technologies beyond simple production tools. One of the main allies in the speed of storage, transport and production in industrial plants are conveyors. Most of the serially developed products are passed through a conveyor belt at the beginning, middle or end of the process. Because of this, it is necessary to effectively monitor the raw material and final parts, to check the conformity of the manufactured product, increasing the quality of the final product, enhancing productivity and reducing costs per loss. In this work, a treadmill support project was implemented, which uses embedded systems and computer vision to monitor the products detected in order to reduce work accidents and increase production by controlling the parts analyzed daily. For the development, a Raspberry Pi was used as a development board and contribution of the OpenCV library, used for the study and programming of computer vision. With the development of this project, it was possible to verify the existence of non-conforming parts, measure the quality of production counting losses and present to the operator numbers that can be useful to improve the process, if applied to a real context. The experiments were carried out in a simulated and controlled context, but they demonstrated that the system is able to perform a compliance check based on the geometric characteristics of certain parts with an average error of 9.17 % in. This result indicates the satisfactory operation for the first version, however, it is important to highlight that the system needs improvement, in order to reduce the error and make its implementation viable.