| Point cloud generation technology based on structured light refers to the technology of point cloud generation by capturing the fringe pattern modulated by the surface of the object and calculating the three-dimensional depth.The technology is in wide demand in non-contact measurement,robotics,medical devices and other fields,where 3D point cloud-based models carry more detail than their original 2D images.For example,cultural relic scanning,through which the three-dimensional model of cultural relic fragments can be reconstructed,can not only see the details of the cultural relic itself,but also more intuitively predict which part of the current cultural relic fragments belong to.The objective of this thesis is to provide a point cloud generation technology that meets the requirements of environment,object and equipment for 3D reconstruction of oral teeth.This thesis focuses on the key problem of this technology: how to complete point cloud computing with high efficiency and high precision.Structured light coding and decoding technology is the core content of point cloud generation technology based on structured light.In this thesis,three innovations are proposed for structured optical codec technology: fringe rearrangement method based on period,periodic fragment rearrangement method and undecoded depth calculation method.There are only five patterns of structured light used in this thesis: three sinusoidal fringe patterns and two mean fringe patterns.However,this picture is not the final projection image.This thesis makes two innovations in the field of coding.The first is the fringe rearrangement method based on the period,which provides a convenient and fast period calculation method.The second is the periodic segment rearrangement method,which solves the fringe discontinuity problem caused by fringe rearrangement on the basis of fringe rearrangement.This thesis explains why the fringe discontinuity problem needs to be solved through an example.In terms of decoding,considering that most fringe projection profilometry takes a long time and has a complicated algorithm in the process of phase unwrapping,this thesis proposes a technique which can calculate depth without decoding.This technology can greatly reduce the amount of computation and simplify the deep computing process,so that the decoding process is more suitable for CUDA development.In the process of decoding,in addition to explaining the principle and technical details,this thesis also discusses in detail some factors affecting the accuracy of structural spot cloud generation technology,such as the number of cycles,and analyzes the accuracy that this method can achieve for random noise.At the end of this thesis,the Geomagic Control software was used to quantitatively measure the accuracy of the point cloud generated in this thesis,and the results showed that the accuracy of the point cloud generated by the method used in this thesis could reach about 0.3mm. |
