Supervisory Control for Automatically 3D Pose Estimation
Download as PDF
DOI: 10.25236/icemit.2019.035
Author(s)
Xiaoli Liu, Xiaoyan Wang
Corresponding Author
Xiaoli Liu
Abstract
Automatically obtaining the pose of 3D object is one of the difficulties in computer vision. In this paper, feedback control theory is used to solve these problems. It is mainly a dynamic target cognitive mechanism with a supervision controller and virtual vision servo controller. Supervisory controller finds the potential matching feature sets according to the feature performance evaluation index, and virtual vision controller gets the possible pose and the integral error under matching hypothesis. The supervisory controller switches in the candidate matches until the integral error is less then specific threshold, then the corrected matchers are found. The augmented reality experiment shows the effectiveness and robustness of the algorithm at the end.
Keywords
Supervisory Control, virtual vision servo, augmented reality, Automatically Registration