Research On Technologies Of Three-Simensional Reconstruction Of Binocular Vision For Additive Manufacturing

The process control of additive and subtractive is used widely in the field of metal additive manufacturing for its advantages in production efficiency,manufacturing accuracy and cost.Among them,how to reconstruct three-dimensional model of additively manufactured parts quickly and accurately is a key technical issue in the process control of additive and subtractive.In this thesis,in view of the demand for three-dimensional reconstruction of molded parts in metal additive manufacturing,several key technologies such as image segmentation,feature extraction,and stereo matching in the three-dimensional reconstruction process based on binocular vision are studied.On this basis,A three-dimensional reconstruction system of metal parts based on surface structured light is developed.First of all,this thesis focuses on the segmentation of binocular image.The threshold method in traditional image segmentation method can't separate the parts in the metal parts image segmentation.To solve this problem,this thesis proposes an image segmentation method based on region growing algorithm and morphological operation.what's more,the validity of this method is verified by the actual image experiment.Then,in view of the defects and limitations of the current feature extraction based on image area in the three-dimensional reconstruction of metal parts,the method of feature extraction under different structure gratings was studied,and a new feature information extraction scheme was designed to solve the difficulty of feature extraction for non-textured metal parts.Next,based on the feature information extraction scheme,a stereo matching algorithm based on coarse matching of Gray code and fine matching of SAD is proposed to realize dense stereo matching of binocular images.On this basis,the neighborhood and matching cost of the matching element are improved and optimized to further improve the matching effect,matching efficiency and anti-interference ability of the stereo matching algorithm.Finally,a three-dimensional reconstruction system of metal parts based on binocular vision is built.The system includes image acquisition,image preprocessing,image segmentation,feature extraction,point cloud matching and curved surface.Through the experiment of three-dimensional contour reconstruction of actual objects,it is verified that the system can realize the function of three-dimensional reconstruction.