Point Cloud Data Processing And Regular Surface Fitting

Because of the the point cloud data got by three-dimensional measurement equipment is too large,has too much noise and so on,some pretreatment process for point cloud data is a must.In addition,for there are a large number of rules surfaces in real life,and such as the parameter with high precision requirement in the process of three-dimensional reconstruction,and the existing methods connot meet the actual needs,so what discussed in this paper is the pretreatment of scattered point cloud data as denoising,data compression and surface fitting method.Main research work includes the following four points:(1)A k neighborhood relations between scattered points has been proposed.Then using the local surface fitting method to calculate curvature of each point.(2)According to the noise characteristics that the noise can be classified into two types :noise in vitro and noise in vivo.K neighborhood denoising algorithm based on statistics to remove noise in vitro,this method first calculates the average neighborhood of each point,then based on the Gaussian curve to set the threshold interval,if the average neighborhood outside of this interval,the point is considered as external noise.Noise in the body,due to its distance from the main body near noise using RANSAC with least squares methods to remove.Experiments show that has more better noise reduction effect.(3)In order to improve the speed of calculation,as far as possible to remove redundant points,retaining points,it is necessary to compress the point cloud.First of all,according to the average curvature shallow areas and mutation,respectively according to the Euclidean distance and the Hausdorff distance of two areas in the regional compression experimentally verified the validity of the method.(4)According to the curvature to distinguish different surface,to complete the identification of surface.Finally,to overcome shortcomings of traditional methods on the measurement accuracy,proposed a total least squares curve fitting method,using the randomly generated rules curved point cloud data and rules of geometry of three dimensionalscanner to test surface point clouds data,experimental data shows that this method works better than conventional least squares fit.