Research On Skeleton Extraction Of 3D Point Clouds Based On Segmentation

Skeleton is an abstract representation of 3D models.It can effectively describe and express the topology feature and shape of the models.With the development of computer graphics and visualization technology,skeleton extraction is widely used in various fields.According to the theory of 3D model segmentation,the model is divided into different regions,then the skeleton for each region is extracted to effectively express the feature of 3D model.In order to address the problem of the skeleton extraction and segmentation for Terracotta Warriors,this thesis did a series of researches.The results of this thesis can be used for shape analysis and recognition retrieval of Terracotta Warriors.The main work is as follows:First,a segmentation algorithm on point clouds was proposed.Aiming at the problem that the traditional segmentation algorithm is prone to over segmentation,in this thesis,the MRF model is applied to the region segmentation of point clouds.To begin with,by using the geometric features of the point cloud models as a prior information,the objective function is determined according to the point clouds properties and the correlation of the neighborhoods.Then,the optimal labeling of the point cloud models is obtained to complete the segmentation of the point clouds.The experimental results show that the algorithm can improve the adaptability of the segmentation process and the accuracy of the segmentation results.Second,this thesis proposes a skeleton extraction algorithm of point clouds based on region segmentation.For the problem of traditional skeleton extraction method easy to cross the region in the dense model of adjacent regions,this algorithm in the segmented region forms the connection points in the segmented region,and compute initial convergence neighborhood scale and extracted the regional skeleton branches,finally complete skeleton branches are obtained by connected connecting points in adjacent regions.The experimental results show that the algorithm reduces the computation and iteration times of global iteration,at the same time,on the basis of the local weighting density to the point clouds,the influence degree of the noise points is reduced,and ensure the correct topology of the skeleton.Besides,by introducing the concept of connection angle and establishing a new connection,the integrity of the skeleton is ensured.Third,in this thesis,the algorithm is mainly applied to the project of "the method of virtual restoration of damaged Terracotta Army based on global optimization".For the skeleton extraction of terracotta point clouds,the skeleton of the original models are obtained,which support the topology analysis and shape matching of digital cultural relics models in the future research.