Research On Virtual Stereo Vision System With A Single Camera And Two Planar Mirrors

With the rapid development of computer aided design and advanced manufacturing technology, relative technologies and applications based on reverse engineering has gradually become a hot research topic in recent years.3D reconstruction, one of the key technologies in reverse engineering, is a research hotspot and challenge in advanced manufacturing technology. It has great significance to reconstruct 3D objects rapidly and effectively.Among the numerous 3D reconstruction methods, stereo vision measuring technology is growing in popularity with its fast speed measuring, high precision reconstruction and strong applicability. It gradually becomes a mainstream of modern 3D reconstruction. Traditional stereo vision measuring systems require two or more cameras, in which the binocular stereo vision measuring system is the representative. However, this kind of system needs to use two cameras separated from a certain distance to simultaneously photograph the object to be reconstructed in different point of views, which may cause increased cost, enlarged structure, measuring inefficiency. Moreover, the system cannot guarantee the synchronism of photograph. In order to avoid the shortcomings mentioned above, a virtual stereo measuring system that just uses a single camera and two planar mirrors is proposed. The proposed system not only combines the advantages of the monocular vision and binocular vision, but also reduces the cost and improves the efficiency. The main content of this paper includes the following parts:Firstly, two virtual cameras are created by using one camera and two planar mirrors. A virtual binocular stereo vision system is built based on the created virtual cameras. The corresponding mathematical model of the system is constructed, and the optimal value ranges of structural parameters are supplied by analyzing the effect of the structural parameters on the measuring field of view.Secondly, camera calibration technique and epipolar geometry model are briefly introduced. An easy and effective mirror plane calibration method is proposed. To calibrate the system, the external parameters between two virtual cameras are obtained by calculating the transformational matrix between the two virtual cameras and the real camera.Finally, the measurement accuracy is enhanced by optimizing the external parameters of two virtual cameras using the Levenberg-Marquardt optimization algorithm, and the effectiveness of the proposed system is validated by the experiments. Meanwhile, the corresponding error analysis of the reconstructed corners of chessboard is conducted, and different level of Gaussian noises is added to the extracted corners of the chessboard to analyze the robustness of the established virtual stereo system.