Normal Estimation and Normal Orientation for Point Cloud Model Based on Improved Local Surface Fitting

The accuracy of normal estimation and normal orientation have great impacts on the processing of point cloud models,such as denoising,registering and surface reconstruction.In terms of normal estimation for3 D point cloud models,firstly the local plane approximation based on PCA method was used to get a preliminary normal estimation.And then the improved Moving Least Squares Surface(MLSS) method was used to get local approximate surface,and thus produce more accurate normals which were resilient to noises.For normal orientation,a new rule of tangential constraint was proposed in the process of normal propagation.Finally,the Poisson surface reconstruction approach was employed to verify the effectiveness of estimated normals.Experimental results show that the accuracy of normal estimation is improved by our method,and smooth models can be obtained as well.Our proposed method can be very useful in the preprocessing of point cloud models.