| With the emergence of the concept of "digital cities," the demand for three-dimensional model reconstruction is increasing.How to process large-scale and spatially complex point cloud data more clearly has become a new challenge.Spectral clustering algorithm has been widely studied due to its strong generalization ability to point cloud datasets and good clustering performance.As the segmentation effect of spectral clustering algorithm depends on the construction of its similarity matrix,traditional spectral clustering algorithms often only consider the spatial distance between sample points during the construction of the similarity matrix.To address this issue,this paper proposes a method for constructing an adaptive similarity matrix based on normal vectors.By combining sparse representation and norm,an adaptive spectral clustering algorithm based on normal vectors(NVASC algorithm)is proposed.The paper presents the iterative optimization process for solving the objective function of the NVASC algorithm and provides convergence analysis.Furthermore,because the spatial complexity required by spectral clustering algorithms is too high,it is difficult for them to achieve the ideal segmentation effect or even realize clustering segmentation when dealing with large-scale datasets.To solve this problem,this paper proposes a representative point-based multi-view weighted spectral clustering algorithm(RPMSC algorithm)based on multi-view clustering ideas.The paper presents the process of solving the objective function of the RPMSC algorithm in three parts and analyzes its convergence.Finally,through instance simulation,the two improved algorithms are applied to the 3D point cloud datasets of chairs and teaching buildings,and the experimental results of the improved algorithms are compared with those of k-means algorithm,FCM algorithm,and SMSC algorithm.The results show that the NVASC algorithm and RPMSC algorithm are both highly practical. |
PDF Full Text Request