Motion and scene interpretation from two views an optical switch

In an attempt to assemble an MEMS optical switch, two views, a microscope view (MV) and a perspective view (PV), are used to observe the scene. This paper presents both the out-of-view reconstruction and motion estimation technologies for the assembly process. For out-of-view reconstruction, a homography capable of transforming feature points and feature lines between an MV image and a CAD model is implemented. The homography between two sequential images is decomposed and factorized for motion estimation. Optical flow has also been explored to provide rough estimation of rotation center and angle. After non-linear optimization, accuracy for estimation of rotation angle and rotation center can reach 0.05 degree and pixel level respectively.