A Multiple-view Based Self-calibration Method For Structured Light System

As a typical active vision system, structured light3D vision system is widely applied in fields such as industry measurement, reverse engineering and the protection of cultural relics. How to build a mathematical model which can display the process of projecting images from2D to3D space coordinates, and adopt a convenient and reasonable visual equipment calibration method are significant research contents.Based on multiple-view projective geometry, this paper puts forward a self-calibration and measurement reconstruction method. In detail, the self-calibration method benefits from theories of projective geometry such as vanishing line, epiolar geometry and infinite homography. The multiple view-based structured light self-calibration method is reasonable and valid as it is derived from the theories above and has been testified in simulated and real situation. Research contents and innovations of the paper are listed as follows:(1) The paper proposes a multiple-view structured light model, in which camera obtains the image of substance being measured in three non-collinear position. The camera centers and the projector model are set to coincide with their optical axis in order to accomplish metric reconstruction and reduce the complexity of self-calibration. In addition, the self-calibration method does not need a target, by which the self-adaptability of the system is greatly improved.(2) The paper puts forward a multiple-view structured light-based self-calibration algorithm, which requires the camera to obtain images in three positions. The vanishing line of the stripe surface of the projector in the first camera coordinate is identified through the homography matrix which is introduced by the stripe surface of the projector model. The intrinsic parameters of camera and projector are obtained from the infinite homography. With the knowledge of epipole point in multiple-view geometry, the relations between extrinsic parameters are identified. Finally, validity of the measurement is testified by simulated and real tests.(3) Structured light planar calibration discussed in the paper has been developed to a comparative mature state with high accuracy. In order to prove the validity of the structured light self-calibration method, the paper adopts planar calibration method as a reference. When metric reconstruction is completed and errors are compared, it concludes that multiple-view self-calibration method can meet the requirements of metric reconstruction, and is easy to operate, highly adaptable and widely applied.Finally, the paper summarizes the limitations and also provides suggestions for further studies. As a highly accurate and quickly reconstructed active vision technique, structured light3D reconstruction technique is directed to improve its real-time performance, systematic robustness and applied range.