Detecção de objetos para auxílio de robôs de serviço hospitalares

• Ano de defesa: 2024
• Autor(a) principal: Silva, Felipe Fabricio Fischer da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Brasil; Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial; UTFPR
• Palavras-chave em Português: Processamento de imagens; Robótica; Visão por computador; Hospitais - Automação; Controle automático; Algorítmos - Desenvolvimento; Image processing; Robotics; Computer vision; Hospitals - Automation; Automatic control; Algorithms - Development; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA::ELETRONICA INDUSTRIAL, SISTEMAS E CONTROLES ELETRONICOS::AUTOMACAO ELETRONICA DE PROCESSOS ELETRICOS E INDUSTRIAIS; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/34367

The objective of this master’s dissertation is to construct a simplified object detection system, using only a camera, for a simulated robot. The motivation to make this project is to aim to improve an autonomous navigation system in hospitals, looking to increase the security and efficiency when transporting medicine to patients. The system aims to enable the robot to detect various objects in a hospital environment, positioning markers equivalent to their position in the robot’s internal map, aiming to improve the precision of autonomous navigation techniques. The ultimate goal is to transfer these techniques to a real robot, capable of delivering medications and documents within the hospital premises, given the lack of security inside this environment and the possibility of making the doctors able to spend more time with the patients. The methodology employed in this study comprises the acquisition of pictures and the development of an image processing model, the creation of an algorithm for distance detection using a single camera, and the placement of markers based on the detected distance and the conversion of quaternions to Euler angles. The image processing approach is validated by confusion matrixes and mAP graphs, to certify the quality of the image detection algorithm, and the distance approach is validated by comparing the achieved results with real ones, verifying the quality of what was developed. The conclusions drawn from the study are favorable, indicating that the robot is capable of generating markers on the map by recognizing objects through the camera, with positions closely approximating reality, obtaining an error margin on the marker of less than half a meter.