Markerless Posture Estimation of Flexible Cord-like Objects by Video Processing with Deep Learning
Mitsuhito Ando, Dainba Ohoba, Kazashi Nakano, Megu GUNJI, Haruo Noma, Hiromi Mochiyama, Ryuma Niiyama · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2023
In this study, purpose is pose estimation of flexible objects. The object is a cord-like flexible object, and moving image analysis using deep learning is used. The method proposed is markerless pose estimation. The movement of flexible objects is captured as a video and labeled on the image. It learns from labeled images and estimates labeling points. A curve is calculated from that point, and the pose is estimated. In this paper, we actually photographed the movement of a flexible cord-like object, and compared the estimation result with the actual photograph for verification. It was shown that a posture similar to the actual posture can be estimated. The maximum tip error was 2 cm.