Recognition of Cloth Target Based on Photograph Modeling and Handling

抄録

Human can recognize easily the different kinds of object. However, it is difficult for robots recognize the objects such as cloth, string, moving objects and etc. This paper is to develop the cloth recognition scheme for visual servoing to apply it in industries where workers recognize clothes and package them to send internet users who have purchased. In this study, the 7-link manipulator equips with two cameras that detect position and orientation (pose) of cloth using 3D-MoS (Move on Sensing). The proposed system introduces a new model generating method for model-based cloth recognition that utilizes a photograph of cloth as a model to recognize the cloth, which enable that the robotic system does not need to prepare a definition of the target cloth as a preparation of the recognition. It has been confirmed that cloth can be recognized by the system through 1000 times recognition experiment.