Research On 3D Denture Image Registration Technology Based On Point Cloud Model

With the advancement of digital technology,it has become common practice to use surgical navigation systems to assist doctors in performing surgery more conveniently and accurately.Registration technology is essential for achieving precise surgical navigation positioning.Through registration technology,it is possible to achieve the transformation relationship between real patient space and virtual image space,to unify the coordinate system of both,and realize the combination of virtual and real.At the moment,it primarily separated surgical navigation registration methods into a marked point and an unmarked point.Among them,the marker-based registration technique has issues with marker falling,blockage,and nonrecognition.This thesis uses the registration method based on unmarked points to circumvent the aforementioned flaws.The following is a summary of the key research findings of this thesis:1)Oral surgery navigation registration technology.In this thesis,key technologies such as surgical navigation registration,multi-modal medical image registration,and point cloud spatial registration introducing and studied within the context of implantology.Following that,this thesis analyzed the entire registration technology process,relevant theories derived,and the basic requirements of registration technology in oral surgery navigation and the technology’s shortcomings summarized,leading to the need for high precision and fast registration algorithm research objectives.2)Acquisition and processing of 3D images.To begin,the real denture three-dimensional image collected in the Point Cloud Library of Point Cloud Library(PCL)is processed through point cloud processing,filtering,and down-sampling data to achieve the real denture point cloud.Meanwhile,the CT and MRI 2D images are fused and reconstructed using the open-source software platform 3D Slicer,and it achieved the virtual denture point cloud through filtering and down-sampling.3)Technology and advancements in point cloud space registration.Rough registration and fine registration are two types of point cloud space registration technology.Iterative Closest Point(ICP)algorithm has shortcomings such as falling into a local optimal solution,a difficult selection of optimal iteration threshold,and a long algorithm operation time.To solve the aforementioned problems,this thesis suggests improvement techniques.To begin,this thesis used the Principal Component Analysis algorithm(PCA)in rough registration,and it introduced the initial correction to solve the problem of the inverted principal axis.The bidirectional K-nearest neighbor search approach,which is not based on the nearest neighbor point pairs,then cuts the running time by exceeding 25.5%.This thesis avoids the point cloud registration problem of the local optimal solution in fine registration thanks to the initial pose consistency.Secondarily,in the fine registration,the error function minimized by minimizing the plane distance between corresponding points,and a K-nearest neighbor search strategy based on K-dimensional Tree(K-d Tree)was used to expedite the search for corresponding point planes.Before searching,the Open Multi-Processing(Open MP)multi-threading framework is used to expedite the point cloud’s normal vector estimation process,reducing the algorithm’s running time.Finally,the spatial registration algorithm’s accuracy,efficiency,and applicability validating through design experiments and software integration.According to the experimental findings,the revised registration algorithm satisfies the accuracy and navigational speed needs of oral surgery.