Adaptive Control of a Camera-Projection System using Vision-Based Feedback

This thesis derives an vision based feedback control strategy for a class of uncertain projector-camera systems that are used to animate two dimensional projected images on complex, three dimensional, articulated target objects. The target object of the robotic system is articulated using an open loop control strategy that generates a desired sequence of target poses that are designed using commercially available geometric modeling software. The ideal or desired image sequences are subsequently rendered in the geometric modeling software using an ideal camera/projector pose and ideal intrinsic parameter camera model.

The rendered imagery from the ideal camera and projector pose are subsequently used to define tracking performance for the feedback control of the camera and projector.

Uncertainty in actuator models of the camera and projector actuator subsystems in this paper includes contributions due to imprecision in camera pose and in intrinsic camera parameters. A feedback control strategy is derived that employs pixel coordinates of multiple tracked feature points in the target image sequence for pose estimation and tracking control problems.

We establish sufficient conditions that guarantee the convergence and asymptotic stability of the pose estimation and tracking control problems for the class of uncertain, nonlinear systems studied in this thesis. Several numerical studies are summarized in the thesis that provide confidence in the derived theoretical results and further suggest robustness of the control strategy for the considered uncertainty class.