Pose estimation for plane based on monocular microscope vision system

摘要

Calibration techniques for microscope vision and pose estimation algorithm for plane are presented based on monocular microscope vision system, which used for visual guiding the manipulation of micro-parts. In contrast to previous approaches, a virtual calibration pattern is constructed by using a high-precision positioning robot with a calibration sphere. Four calibration algorithms based on single parallel plane for microscope vision system specifically are discussed. And through comparing, the adequate calibration approach for microscope vision is selected. Then a positioning algorithm is applied for estimating the pose of the plane in real-time. And the initial value of the pose is estimated by the linear methods base on PnP algorithm. Then the precision value of pose is optimized by the orthogonal iteration algorithm. The validity and accuracy of the proposed methods are tested by experiments. The projection error of the calibration is 0.8 pixels and the average error in distance from the measured position to the actual position is 31µm.