Summary: | 碩士 === 國立成功大學 === 機械工程學系 === 105 === The objective of this thesis is to implement augment reality in manual micro-assembly operation through 3D model-based virtual camera calibration and visual servo, where “3D” means the proposed algorithms take object 3D coordinates in virtual environment as control points. First of all, apply virtual camera calibration by the linear model, non-linear model and stepwise method to adjust camera pose, which can let the appearance of virtual objects on the screen is the same as the scene captured by real camera. Next, use the relationship between object velocity and its reprojection point velocity on image plane to build the virtual visual servo model, which can be used to derive the control law to let the virtual object track an assembled part dynamically. If the assembled part carried by linear stage rotates during transportation, pose of the virtual object can be adjusted by the proposed control law. Moreover, the feature points used in calibration and visual servo are obtained by image processing and curve fitting. When executing assembly operation, the invisible hole inside the assembled part can be revealed by overlapping virtual object on real image and utilizing face culling method, which can assist the operator in locating position of the hole and increasing the success rate in assembly mission.
|