3D Reconstruction Based On Nonparallel Two-eye System

As the rapid development of the modern manufacture industry,traditional measure method limits productivity and the precision of the products.High accuracy 3D data extracting of industry products,such as appearance of the mould,engine outer cover,automobile appearance,has great practical significance in designing products data,testing manufacture results,comparing products with original data and supplying data to the subsequent steps.So how to obtain high precise and abundant 3D data of industry products fast is urgently to be solved.During the last few years,computer stereo vision has been greatly developed and applied in 3D measurement field.Compared with the traditional contact method,measurement technology based on computer stereo vision has many advantages:non-destructive,high speed and high precision.Based on the program,comparing and analyzing large steel boards' 3D data,this paper studied the whole process of 3D reconstruction,including calibration,image correction,feature extraction and 3D data reconstruction.It is an important application of computer stereo vision. Various methods for 3D reconstruction can be found from the literature.However,surprisingly most of them are always concentrated on some procedure of the whole 3D reconstruction and the way used is always feature extraction,feature register and 3D reconstruction.In this paper, feature extraction can reach a sub-pixel precision,and illumined by R.Balasubramanian.the theory of quadratic curves for 3D reconstruction is used in the program.Here,just the correspondences of curves are used,that avoids the disadvantage of points register.Also,this paper proposed a new cross center identification method and improved many procedures of the whole process.Finally,the results prove that the process is feasible.