| With the development of computer-aided design technology, reverseengineering technology which can construct digital model via physical modelis more and more widely used. However, in the process of three-dimensionalscanning device to get point cloud data, due to human factors andmeasurement environment factors or the defect of the scanning device itself,and many other factors, make the point cloud are subjected to noise pollutionin some way. Therefore, before the digital geometry processing andapplication of the point cloud data, we must carry out filtering noise reductionprocessing on the point cloud. The goal is to reduce the noise and keep thetopological characteristics and geometrical characteristics of the cloud modelsample surface the premise of effectively eliminate noise and reconstruction ofthe original smooth surface.In this paper, the three-dimensional scattered point cloud data noisereduction technology is studied, the main innovation points of this method isas follows:1) Before the point cloud data are multilateral smooth, firstly eliminatingoutliers with the k nearest neighbor search method based on density in pointcloud, the outliers are removed to reduce the original point cloud data, andreduces the time complexity of the subsequent noise reduction processing,improve the efficiency and accuracy.2) Give full consideration to the effect of the pretreatment for samplingpoints and its optimum in the neighborhood of the Euclidean distance in thenoise reduction effect, redefines the noise reduction algorithm of theweighting function space, makes the farther distance pretreatment samplepoint the smaller the contribution to the noise reduction, increasing the pointcloud of multilateral filtering noise reduction, improve the efficiency.3) Considering the characteristics and noise belong to high frequencysignal, the original algorithm can put some characteristics of mistaken fornoise condition. When the curvature is bigger if the optimal neighborhoodrange is small, then the sampling points for the noise probability is very big, based on the curvature function weighted processing, the accuracy ofmultilateral filtering noise reduction algorithm is improved.In this paper, the improvement of multilateral filter algorithm has carriedon the experiment. The experiments show that the improved algorithm ofscattered point cloud data can not only effectively noise reduction processing,and in maintaining the feature information, and prevent the volume shrinkagealso show some advantage. In this paper, the experiment is only for simplepoint cloud data and the gaussian noise, geometric features were notparticularly sharp and complex point cloud. |
