A method for 3D measurement and reconstruction for active vision

In this paper, we present a new method for 3D measurement and reconstruction for an active vision system. The vision system consists of a light pattern projector and a camera. With initial calibration of the other components in the system, the camera is allowed to change its internal and external parameters during a measurement task. This gives the system the ability to adapt to its environment or task. With this method, the image-to-world transformation is recovered on-line. The 3D scene can then be reconstructed via this transformation. Compared with other existing methods, our approach has the following features: (1) the CCD camera is allowed to undergo an unconstrained motion or change in focus or any of its parameters. In fact, no prior knowledge of the camera parameters is needed by our approach; (2) the computation cost is lower than the traditional method; (3) the method is linear in computation as only a set of linear equations needs to be solved. As a result, the inherent problems existing in nonlinear system, such as divergence or local convergence, are overcome. Simulation and real experiments have been conducted using our active vision system. Our method proves to be valid and the measurement and reconstruction results turn out to be satisfactory.