Hybrid Feature Based 6D Pose Tracking under Binocular Vision for Automated Micro-assembly

Abstract

6D tracking is important in robotic based assembly of the MEMS. This paper presents a 6D pose tracking method using hybrid feature for automated micro-assembly. Firstly, the micro-assembly system which mainly comprises of a parallel positioning robot that can largely rotate around three axes, a serial 6-axis robot for manipulation, and a binocular microvision modular is built. To realize tracking of the 6D pose under large rotation, a hybrid feature based tracking scheme that can be used under multi-view is developed. Experiments are conducted to verify the performance of the proposed method. The results show that the tracking performance can be effectively improved by simultaneously using the point feature and line feature. Besides, the averaged translation error and rotation error are smaller than 0.031 mm and 0.3 degree, respectively.