Point Cloud Feature Recognition Method Based On Local Sample Frame Constraint And Its Application

The feature recognition of point cloud data is a key issue in the field of reverse engineering.It has important guiding significance for subsequent processes such as point cloud segmentation,point cloud reduction and surface reconstruction.Because the existing methods have redundant noise and feature point omission in the extraction of point cloud feature area samples,the accuracy of the feature point set is impaired.In order to improve the point cloud feature recognition effect,this paper focuses on the optimization of the surface local sample.The point cloud feature recognition method and the application of feature recognition results in the surface reconstruction process are studied.The main progress achieved is as follows:(1)A point cloud feature recognition method based on local central axis feature constraints is proposed.Based on the principle of the central axis,the distribution characteristics of each central axis point are analyzed.According to the difference of the shape and position distribution of the central axis in the different shape of the curved surface,it is divided into three categories:out-of-group axis,shape-maintaining axis and limit axis.Then,based on the Voronoi partition,the point cloud poles are obtained,and the characteristics of the central axis shape are inferred by means of mean clustering,plane fitting,principal component analysis,etc.of the local region pole set,and then the attributes of the target sample are distinguished,and the feature area is completed.Identification of samples.The method can be used to solve the problems of redundant noise and feature omission in the current mainstream methods in identifying point cloud features.(2)A local sample optimization method based on the central axis is proposed.The central axis is fitted by the Voronoi pole in the vicinity of the given sample,and is filtered according to the shape distribution characteristics of the central axis,from which the central axis of the limit in the European neighborhood of the given sample is retained,based on the required limit.The isolation effect of the center axis on the curved surfaces on both sides eliminates the noise in the Euclidean neighborhood of the given sample point,and completes the optimization of the surface sample.The method can be used to acquire the geodetic neighborhood data of any given sample point in the point cloud,and effectively avoid the interference of the noise on the topographical characteristics of the local sample.(3)A point cloud feature recognition method based on local sample frame constraints is proposed.Based on the local geodetic neighborhood data of the surface sampling points,the local sample frame is constructed by the steps of normal estimation and quasi-common method,and the local samples of the curved surface are quantized in different discrete directions according to the angle of each frame line.The morphological features are used to accurately distinguish the sample points in the point cloud data that are located in the smooth region,the edge region,the sharp corner region,and the boundary region.The method can significantly improve the recognition effect of the thin-walled area samples,and is suitable for the identification of the boundary area samples,further improving the point cloud feature recognition effect.(4)A method of surface topology reconstruction based on feature recognition is proposed.Firstly,the classical Curst reconstruction method is used to reconstruct the surface of the point cloud data.The point cloud feature recognition method with the central axis constraint is used to obtain the sample points of the sharp edge and its adjacent area,and the incident patch is removed as the defect patch.Get a smooth area mesh surface.Then,using the point cloud feature recognition method of local sample frame constraint,the sample points located at the feature line and the feature boundary are extracted from the rough feature point set,and the whole surface topology reconstruction is completed by the steps of feature line reconstruction,smooth region incremental expansion and feature line stitching.Obtain a two-dimensional directional manifold mesh that is interpolated to the target sample point set.The method can significantly improve the reconstruction effect of the sharp feature area of the point cloud,effectively eliminate the reconstruction defects such as non-manifold patches and holes,and avoid the appearance of the narrow and long patches.