Research On Three Dimensional Reconstruction Algorithm Of Denture Model

Reverse engineering is developed and applied in the field of biomedicalscience constantly. CAD technology is used to oral medicine quickly. With theadvantages of high precision of denture and short time of production cycle, etc, itwins good graces from doctors and clients and will be the inevitable trend of oralmedicine. Although the denture CAD system has been designed overseas, itrequests the point data of denture too strictly and expensively to be appliedwidely. The three dimensional point data of denture is regarded as research objectin the paper. The basic theories and methods are researched on the denture pointdata with only position information, namely pretreatment, surface reconstructionand post-processing reconstruction in three dimensional reconstructiontechniques.According to the denture data which have only position information and lacktopological relations, this paper proposes a dynamic k-nearest neighbor queryalgorithm based on Morton order. Three dimensional data are divided in spacefirstly and stored in Octree structure. Morton code for every point is generated byMorton order. Three dimensional data can be changed to one dimensional data.And then a parallel fast sort algorithm based on Euclidean distance is adopted tomake the Morton code in order and stored in one dimensional linear chain.Finally, Morton code can be located corresponding point clouds directly byvisiting the linear chain, avoiding the complexity to search in subspace even toneighbor subspace. The experimental results show that the proposed algorithm issuperior and accutate.The existing surface reconstruction methods are excessively dependent on thenormal, so fuzzy normal estimation algorithm is proposed for point clouds ofdenture model. The point clouds are classified by fuzzy logical system accordingto point clouds density and curvature change. The smooth area of crown, thethin-plate one of cusp or teeth edge, and sharp one of molar occlusion surface are divided. PCA algorithm is used to smooth area, the checker one is used to thin-plate area, and the attach point one is used to sharp area. The output of fuzzylogic system determines the tooth point cloud data should adopt thecorresponding normal vector estimation method. The two latter methods arecomplex but accurate. Because the area including thin-plate and sharp feature isless than smooth area in denture, the two latter algorithms are used low frequency.On the contrary, PCA algorithm is used frequently. The experimental resultsshow that normal estimation of denture points is efficient and accurate.Due to limitation on the number of mass cloud points, a progressive implicitsurface reconstruction is proposed. The selection of offset point is developedbased on the center point of sub-cube to decrease the number of radial basisfunction needed during the denture reconstruction. A simple coefficientcalculation of implicit function is proposed based on Schur compensation, thecoefficient matrix is partitioned into four blocks, and only one of the four blocksneeds to be updated during the surface fitting. In this way, the calculation reducesan order of magnitude for mass point clouds. The local surface reconstructionalgorithm is proposed in Octree sub-cube. The denture model and tooth modelare fitted by this algorithm. The results show that it is feasible and its advantagefor saving time is obviously compared with other methods.Aiming at the irregular of sharp feature distribution in denture and poor effectof hole-filling, the definition of sharpness-filter is proposed. The vertexes of holeare divided into two classes by computing sharpness of vertex on the mesh ofhole. One class with small sharpness is repaired and the initial model isconformed by implicit surface. Another with large sharpness such as canine ortooth fossa is used by the algorithm proposed in this paper based on implicitsurface. The function of proposed algorithm is to recover the feature of surface,which allows the sharp feature to be extended to the surface patch inside of thehole and the parameter of sharpness is adjusted according to the partial occlusionarea of teeth. At last, this paper proposes a simple surface smoothing method ofdenture model. The method is equivalent to transporting the scalar implicitfunction by solving the system of coupled non-linear ordinary differentialequations (ODE) and the propagation of the surface can be performed by usingthe approximate solution of an explicit scheme. Theoretical analysis and experiment show that the hole-filling algorithm and surface smoothing are simpleand useful for denture, which can repair and fill the hole of denture effectivelyand recover the shape of original model.