Research On 3D Scanning Technology Based On Monocular Vision

With the progress and development of science,people are also looking forward to the three-dimensional structure model of the surrounding objects.People can now scan objects through many devices for contact type or non-contact type,to achieve the reconstruction of three-dimensional model,such as laser,structured light,radar and so on,while 3D-based visual reconstruction based on vision is more economical.However,the problem of the slow speed of the three-dimensional scanning reconstruction based on the monocular vision that seriously affected its convenience and scope of use.In this paper,the related technology of 3D scanning based on monocular vision is studied.The main contents and achievements are as below:A monocular 3D scanning method based on packet reconstruction has been designed.On the basis of the research of SFM algorithm,it has been improved and the packet reconstruction framework has been designed.Firstly,the image data is divided into N groups,each group extracts the SURF feature from the image and reconstructs it with SFM.Secondly,the results of the reconstruction of each group on the scale of registration and fast splicing access to the overall point cloud model.Thirdly,the block registration strategy is adopted to weaken the effect of the accumulation error of point cloud matching.Experiments show that this method can greatly enhance the speed of 3D scanning reconstruction under the condition of without losing accuracy,and a larger number of images is more suitable for packet reconstruction.The final experiment verifies the feasibility of this method.A fast registration algorithm of three-step point cloud based on NDT and ICP is proposed.Considering the registration accuracy of ICP algorithm and the registration speed of NDT algorithm,the three-step registration strategy is adopted to speed up the registration speed of point cloud.Firstly,the NDT algorithm is used on the point cloud rough registration,then the distance and posture between the two points is adjusted quickly.Secondly,the ICP algorithm is used to fine-tune the set of points after coarse matching,and adjust the position and pose of the cloud.Thirdly,ICP algorithm is used again to precisely register the set of point cloud.Through the experimental verification,the algorithm proposed by this paper can effectively shortens the registration time without losing the registration accuracy and larger increases the speed of three-dimensional scan reconstruction,especially for a large number of point cloud or distant point cloud.