Study On Anisotropic Denoising And Preserved Edge Simplification Of 3D Point Cloud Data Of Aeroengine Blade

As one of the core components of aeroengine,blade has become the key repair object in aviation maintenance industry because of its large number,complex shape,high manufacturing cost and easy damage.Reverse reconstruction of blade digital model is an important part of blade automatic welding repair technology.High-precision blade three-dimensional point cloud data is the premise and foundation of blade reverse reconstruction.However,the measured large-scale point cloud data of blade contain noise because of the defect of measuring equipment and the influence of measuring environment,and the accuracy of point cloud data is reduced.In addition,the large-scale point cloud data makes the time of point cloud surface reconstruction becomes larger,and makes the efficiency of point cloud surface reconstruction becomes lower.Therefore,it is important to study on high-precision point cloud denoising technology and simplification technology for welding repair of aeronautical blade.In the aspect of point cloud denoising,the traditional isotropic denoising algorithms of point cloud model easily lead to over-smoothing and loss of local feature of the model,and the model is distortion after denoising.In the aspect of point cloud simplifying,the traditional simpliying algorithm of point cloud easily lead to the defect of edge feature of the simplified model because of over-deletion of feature point.To solve the above two problems,the anisotropic denoising algorithm and the edge preserving simplification algorithm of point cloud were studied.A 3D scattered point cloud smoothing denoising algorithm was proposed based on anisotropic diffusion filtering through analysising the local geometric feature information of the sampling point.The tensor matrix of the sampling point was obtained by tensor voting algorithm,and the eigenvalues and eigenvectors of the tensor matrix were solved.To adjust adaptively the diffusion rate of the different characteristic direction,the eigenvalues of diffusion tensor were designed based on the eigenvalues of the tensor matrix.The anisotropic denoising of point cloud model was realized by combining the reconstructed diffusion tensor with the three-dimensional anisotropic diffusion filter equation.The experimemtal results of different point cloud models with noise showed that the feature information of the originally model was preserved effectively and the noise from the point cloud was removed.By estimating the normal and curvature of point cloud,an edge point extraction algorithm was proposed based on normal and curvature of point cloud,which could extract edge data points accurately.On this basis,a preserving edge featur point cloud simplification algorithm was proposed by recognizing the edge feature point of model.The point cloud data was divided into edge data point and non-edge data point,and all edge data point was preserved to prevent the loss of edge contour.The hypervoxel clustering algorithm was used to cluster the non-edge point,and the principle that the cluster center replaces the whole cluster was used to simplify the non-edge data point.The simplified point cloud model was obtained by cambining edge point with simplified non-edge point.The results of the simplification experiment of different point cloud model showed that the proposed point cloud reduction algorithm not only realized redundant data point reduction,but also ensured that the edge contour of the point cloud model was not defected.The experimental analysis of the measured point cloud of aeroengine blade was carried out by using the point cloud denoising algorithm and simplification algorithm proposed in this paper,and the point cloud model could retain the sharp feature of blade edge with equal curvature after denoising and simplification.The mean deviation of the point cloud model of blade is 0.133 mm,and the results meet the requirement of welding technology.