Research On Technology Of3D Reconstruction Of Rotational Image Sequences Based On Microscopic Vision

As a new kind of precision measurement technology, microscopic vision technology can help people understand and change the world from the microscopic view, which has great significance and research value of the popularization and industrialization of micro-electro-mechanical systems (MEMS). In addition,3D reconstruction is the core of the microscopic vision technology, and the rapid development of3D reconstruction will bring about revolutionary breakthroughs of microscopic vision technology.With rotating platform and monocular optical microscope, the goal of this paper is to achieve a new microscopic vision system, which is capable of fast3D reconstruction of MEMS components based on image sequences (video) and applied in the precision measurement of MEMS components. The main work and achievements of this paper are as follows:1) An identification and location algorithm of the chessboard-grid corners automatically under microscopic vision system is proposed. With the help of Zhang’s plane calibration, an automation process is completed to calibrate camera parameters of microscopic vision system when rotating platform is in the initial position.2) A new calibration method of camera extrinsic parameters is proposed for image sequences, in which chessboard-like patterns are rotated in a plane. In this method, the rotating geometric model is established and the rotating axis is calibrated. This method can solve the problem that Zhang’s plane calibration method is not used to calibrate the camera parameters of images in the same plan.3) A new mechanical error evaluation method of rotating platform is proposed based on rotating geometric model and reprojection method, and then an error compensation model of rotating platform is built for the preprocessing of microscopic image sequences (video), which will enhance the accuracy of rotational stereo vision3D reconstruction in later.4) A new3D reconstruction method is proposed combining with silhouettes reconstruction and rotational stereo reconstruction, which uses known camera parameters and the contour extraction of target object in microscopic image sequences through Canny edge detection algorithm and mathematical morphology method. In this method, a sparse point cloud of the3D surface structure is acquired by rotational stereo reconstruction, which as a boundary constraint of visual hull acquired by silhouette reconstruction, and then a dense point cloud of the3D surface structure is reconstructioned.By analyzing the results of3D reconstruction of the MEMS components, it is shown that this microscopic vision system has the advantages of quick reconstruction, simple operation, low requirements of experimental environment and low cost of equipment. What’s more, the reconstruction accuracy of this system has nearly reached the performance of3D Super Depth Digital microscope, which is one of the most common and mature microscopic vision system currently. In a word,3D reconstruction of the microscopic vision has a very high research value, the microscopic vision system and associated algorithms proposed by this paper has broad application prospects.