Abstract: Aiming at the problems of high computational and complex image calibration process, the image preprocessing, image information extraction and matching were studied. A sequence image reconstruction model method was proposed. Based on the principle of monocular vision reconstruction, the SIFT algorithm was used to extract and match the feature point set of the image, and the singular value decomposition of the basic matrix was performed to obtain the camera matrix. The selfcalibration algorithm of the absolute dual quadric surface was introduced to automatically establish the object coordinate system. The model was restored according to the principle of back projection. A threedimensional reconstruction experimental platform for impeller parts was built, and the mesh model reconstruction experiment was carried out. The results indicate that the proposed method can accurately reconstruct the 3D model of the part. And it can realize highprecision and highefficiency automatic 3D inverse modeling. A reliable model acquisition method to upstream technical support for 3D printing processing is provided.
Key words: 3D reconstruction; automatic camera calibration; scaleinvariant feature transform（SIFT）algorithm; 3D printing