Search

In this paper, we propose a vision-based method to estimate the pose (position and orientation) of the palm. Two sub-problems were solved in this study. One is to obtain the correspondence of the four coplanar mark points on the target object. The other is to estimate the position and orientation of the palm in the camera coordinate system. The position and orientation of the palm in the camera coordinate system can be derived by the geometric relation, which can be transformed the pose estimation problem into a Perspective-n-Point (PnP) problem. Dog-Leg (DL) algorithm is adopted to solve the PnP problem to get a higher processing speed. An experiment is conducted to verify the effectiveness of the proposed methods. The experimental results demonstrate that proposed method can obtain the position and orientation accurately and the DL algorithm can be used to solve the P4P problem faster than an existing algorithm.

Particle filters are being widely used in tracking single object for its unique advantages. According to former studies, particle filters can solve the tracking problems in the circumstances of nonlinear and non-Gaussian observation. In this paper, we propose a robot teaching method based on multi-objective particle filter for motion tracking, in which targets can be recognized by the value of H in HSV color space. After the targets are recognized, particle filtering is applied to achieve motion tracking. PnP algorithm is another essential part of this method, which can obtain translation and rotation matrix of the moving objects. All these pose information is sent to robot to reproduce the targets’ motions. The experiment in this paper is divided into three parts, firstly, a single robot arm achieve the linear motion along three-axis respectively, and then move along a particular trajectory. Finally, to achieve both arms tracking hands movements at the same time. When the single target is tracked, the experimental results are better compared with track hands simultaneously.