Instituições de ensino: Universidade Federal do Rio Grande do Sul, Universidade Federal do Rio Grande do Sul, Universidade Federal do Rio Grande do Sul, Universidade Federal do Rio Grande do Sul
Gostou? Então compartilhe:
This paper presents a computer vision system capable, through the capture and processing of digital images, to identify the position and orientation of several parts arranged randomly in the same surface, in the industrial robot workspace. The image is captured by a camera and processed, where the thresholding is performed in order to obtain a final image with the parts segmented and identified. The center of mass of each part are then defined, and an artificial neural network is applied to determine their orientations. The neural network has a multi-layer feed forward architecture with one hidden layer, and was previously trained with several images of similar parts as those used for the tests. The coordinates obtained can be sent to the controller of a robotic manipulator, which would collect these parts and transfer them to another location previously defined. The proposed method was tested using an image set obtained in order to simulate real situation in an industrial environment, with multiple parts randomly arranged in a surface.
Usuários que poderão acessar os artigos finais: Membros ABCM e frequentadores do Cobef 2019. Os frequentadores do COBEF deverão logar no site da ABCM com as mesmas credenciais do sistema de submissão de arquivos.
Fique por dentro de todas as novidades, eventos, cursos, conteúdos exclusivos e muito mais.