Research On The Structuralizing Method Of Geometric Features For Point Cloud Models

Geometric feature lines have the ability to express the contour information of 3D models.The extraction of geometric feature lines is helpful to obtain the theoretical significance and practical value of surface sampling.In the point cloud model,each point comes from the discrete sampling of the target surface,which has the characteristics of scattered distribution,lack of topological connection,typical disorder and disorganization.At the same time,the complexity of 3D shape and the amount of data make structure support difficult for point cloud analysis and processing.This paper focuses on the problem of geometric feature structuralizing of focus cloud model and proposes a feature line construction method based on geometric feature detection operator.Firstly,the local geometric feature points(including concave,convex and surface boundary)of the 3D point cloud model are obtained by using the nearest neighbor reweighted local centroid operator.The feature point set is refined.Secondly,the detected feature points are connected into feature lines to obtain the intersection points of multiple lines as model corners.The feature line segment is preliminarily divided according to corner points.Then,the control points of each feature curve are obtained by iterative Bézier curve algorithm,which further divides and fits the curves with larger fitting errors.Finally,the control point set of all feature lines can be stored structurally.In conclusion,the main work and contributions of our work can be concisely stated as follows:(1)It is achieved that the corner detection of geometric features on object-level model.By connecting feature points and extending them linearly,the intersection of multiple feature lines can be found.The degree of curvature variation in the neighborhood sphere is calculated at each feature point.The point with the maximum local variation is selected as a corner point.(2)A metric model is constructed to divide the feature curves.Each point on the curve is comprehensively considered to divide the curve,including geometric deflection angle,nearest neighbor distance,degree of curve bending,length of each curve,etc.(3)The regular expression method of geometric feature set of point cloud model,namely structuralizing feature curve,is proposed.The iterative Bézier curve algorithm is used to obtain the control points of each feature line,which can store the control points of each curve in a structured way according to a specific format.By using a set of control points,all feature curves can be generated by explicit interpolation.Through experimenting on object-level and scene-level point cloud models,the scattered feature points are effectively organized into regular feature curves,which uses structuralizing feature curves to abstract the point cloud model.The structure of geometric features of point cloud model can provide technical support for 3D modeling,scene segmentation and object retrieval.