Research On Feature Preserving Repair Of Mechanical Workpieces 3D Point Clouds Based On Analyical Tensor Voting

With the rapid development of China's manufacturing industry,the demand of mechanical workpieces with high-precision and high-quality is increasing and the traditional contact measurement can no longer meet the industrial requirements.As a new non-contact measurement method,Three-dimensional measurement technology is favored by all walks of life.At the same time,cloud data processing technology has become a popular technical means.However,for many reasons such as occlusion and limitation of scanners,the point cloud model of original object may be incomplete,and form different holes.Obviously,these holes influence many subsequent operations.After the study of the existing hole filling technology,this paper presents a novel feature preserving holes filling method based on analytical tensor voting.The major contributions of this paper are as follow:1.The fundamentals of traditional tensor voting algorithm are introduced,and the reasons for computational complexity and low efficiency are analyzed.Then in order to overcome these shortcomings,we present the analytical tensor voting algorithm and thoroughly solves above problems by using the new voting mechanism and obtaining the analytical solution of tensor voting formula.Finally,by using the analytical tensor voting algorithm,the feature extraction of point cloud data is realized and the accuracy and efficiency of this method are verified,which lays the foundation for hole filling.2.A method of multi-criterion hole identification based on analytical tensor voting is proposed.This method firstly analyzes three kinds of hole identification methods,including the maximum angle measurement,half-disc measurement and tensor voting measurement.Then according to their characteristics,the method of multi-criterion hole identification based on analytical tensor voting is proposed by way of the weighted summation.Finally,the experimental results show that this method can accurately identify various kinds of holes,including the hole with sharp feature missing.3.A novel feature preserving holes filling method based on analytical tensor voting is proposed.Firstly,the characteristics of hole are analyzed,and according to results,theholes are categorized into two types,featured holes and non-featured holes.Then the SingleSubVoxelCmarch algorithm is used to restore the missing feature points of the incomplete feature curves.Accordingly original holes were divided into non-featured sub-holes.Finally,all non-featured holes are filled by the tensor voting.Experimental results show that this method can accurately and effectively fill the hole and restore the original feature of model at the same time.4.The experiment of real point cloud data acquisition is designed and simulated,and the hole filling method presented in this paper is verified by the real point cloud data.