(AIMS 2021) Manipulator posture estimation method based on multi-eye vision and key point detection

Abstract: Image recognition is combined with multi-eye stereo vision technology to estimate the pose of the robot arm which can improve the flexibility of the robot arm control and reduce the immobilization of the control mode. Multiple cameras illuminate the robotic arm to form a multi-eye vision model. At the same time, the HRNet key point detection model detects the key points of the robotic arm in the picture, establishes a space line equation set, and obtains the coordinates of the key points of the robotic arm in space, so as to realize the pose estimation. The experimental results show that when the key points of the HRNet model are detected correctly, the pose of the robotic arm can be estimated well. The method has research significance and practical value.

Authors: Aqing Wang and Cheng Cai (Shanghai Dianji University, China)

Email: wangaq126@163.com, cheneychengcai@163.com