Feature Extraction Algorithm Of The Point Cloud Data And Its Application Based Normal Vector Regional Clustering

The rapid development of 3D laser scanning technology makes it possible to obtain high-precision surface models of objects in actual production and life for it can accurately and effectively collect the surface data of the various objects.However,in the process of 3D model reconstruction based on point cloud data,the integrity and accuracy of model feature extraction become the main factors affecting model reconstruction,which is also the basis of subsequent work such as model reconstruction.Therefore,the feature extraction of point cloud data has become a research hotspot in point cloud data processing.In the field of urban and rural development planning,with the advent of the era of smart cities and digital cities,modeling the entire outdoor environment has become an inevitable choice.In addition,the building is a landmark in the modern city.Therefore,how to effectively extract the features of various target point clouds in outdoor scenes dominated by buildings is an important issue in the construction of digital cities.In view of the above analysis,a method of extracting feature lines based on normal vector region clustering was proposed,and the algorithm was applied to extract the features of some typical ground objects such as buildings,vehicles,and benches in outdoor scenes.Furthermore,before the feature extraction of outdoor scene objects,the semantic division of the outdoor scene is carried out by combining cloth simulation filtering algorithm and density clustering method.The main research contents and results are as follows:(1)A method of extracting feature lines based on normal vector region clustering was proposed to solve the problem of incomplete extraction of transition lines and detail feature lines of scattered point cloud models in reverse engineering.Principal component analysis(PCA)of adaptive neighborhood was used to estimate the normal vector of the model,and the firefly algorithm(FA)optimization fuzzy C-means(FCM)algorithm was introduced to cluster the normal vectors for realizing the effective segmentation of the model.The candidate feature points are extracted from the boundary points of each block by constructing the rule of eliminating and merging the point sets,and then the feature points are extracted based on the principal axis direction of the local neighborhood.Four geometric models with different complexity in the field of reverse engineering are selected for experiments.(2)The features of typical objects in outdoor scenes are extracted through the algorithm proposed in this paper.First,the ground points in the original point cloud data set were eliminated by the cloth simulated filtering algorithm,and the density clustering method with adaptive parameters is used to realize the division of different objects in the non-ground points.Then,three typical features of buildings,vehicles,and benches in outdoor scenes are taken,and the features of these three features are identified by the normal vector regional clustering algorithm.(3)The experimental results of four geometric models in the field of reverse engineering show that: the algorithm has good adaptability and accuracy,can extract sharp features and detail features from the point cloud model,meanwhile getting the transition features;The test results of three types of typical ground features in outdoor scenes show that: the proposed algorithm can be applied to extract the features of outdoor scene features,and it can effectively extract the main features of outdoor objects point cloud models.Besides,based on ensuring the density of the point cloud,the proposed algorithm also has a good ability to extract the detail features and transition features of the point cloud model.