Research And Application Of Fuzzy Clustering Segmentation Of 3D Point Cloud Data

With the development of human information age,human's cognition of society is deepening,and the demand of establishing real scene model is also increasing.In digital map and military and other related fields,3D model of real scene is widely used.In building a real scene model,the key points are the point cloud data acquisition,point cloud data processing and reconstruction.3D point cloud data generally contain coordinate information,color information and depth information.In the preprocessing of 3D point cloud data,the clustering segmentation of point cloud data is an important content,and it is the foundation and premise of 3D reconstruction.The clustering segmentation results of point cloud data affect the effect of 3D reconstruction.Therefore,the segmentation algorithm of 3D point cloud data is also in continuous research and exploration.Due to the complex spatial structure of point cloud data and the large amount of data,the existing algorithm can only cluster and segment the single point cloud data,and it is easy to be affected by various factors,which results in a long time of clustering and the clustering results can't meet the requirements.This paper introduces the research status of 3D point cloud data at home and abroad,the basic knowledge of fuzzy clustering analysis,the classification of fuzzy clustering algorithm and some basic methods of point cloud data preprocessing,and analyzes the advantages and disadvantages of some clustering algorithms.Fuzzy c-means clustering algorithm(FCM)is easy to be affected by parameters and noise,and FCM algorithm does not consider the spatial information of point cloud data,resulting in inaccurate clustering segmentation results.Firstly,the point cloud is divided into data.Secondly,the method of discrete point plane fitting is used to fit the divided point cloud and calculate its normal vector.Then,the objective function of fuzzy c-means algorithm(FCM algorithm)is weighted by using the information of point cloud normal vector and cluster center normal vector.The normal vector weighted fuzzy c-means algorithm(NFCM algorithm)is proposed and its convergence is proved.Finally,NFCM algorithm is used for clustering segmentation of building point cloud and palm point cloud.The astringency and feasibility of NFCM algorithm are verified by clustering segmentation results,iterative convergence speed and partition coefficient of NFCM algorithm and FCM algorithm.