Research And Application Of 3D Model Segmentation And Registration

With the development of 3D scanning technology,more and more 3D models can be obtained.The processing of 3D models has gradually become a hot topic.Segmentation and registration of 3D models are important steps in 3D model processing,and are widely applied in many fields,such as cultural relic protection,medicine,reverse engineering,engineering design,animation and so on.However,the segmentation of 3D models still exists the problems of segmentation meaningless or over segmentation,and the registration also has long time and inaccurate results.Aiming at these problems,this paper takes Terracotta Army debris data and 3D craniofacial data as research objects,and puts forward some improvement strategies for segmentation and registration of 3D models.The progress of the main research work includes:1.A 3D required surface recognition method based on minimum spanning tree and pruning strategy is proposed.First,we use integral invariants to extract feature segmentation points.Secondly,minimum spanning trees are constructed and combine it with pruning strategy to form closed boundary lines to construct feature regions.Finally,the required surface of three dimensional craniofacial face or Terracotta Army fragmentation surface are extracted and identified.The experimental results show that the method can effectively complete the requirement face recognition of the 3D model.2.A 3D model segmentation algorithm based on multiple random walk is improved.Based on the probability distribution of all agents in the 3D model,the algorithm is designed to reboot the rules between multiple agents.These agents traverse the 3D model according to the transformation matrix,and integrate all the surfaces into meaningful areas according to the shape characteristics.Experimental results show that the method does not rely on the initial proxy points,and solves the problem of over segmentation and segmentation meaningless.It is also applicable for complex 3D models.3.Aiming at the problem that the iterative closest point registration algorithm takes a long time and there is excessive rotation in registration process,this paper proposes a model registration method combining iterative factor with rotation angle.First,we use the segmentation algorithm to find the feature area of the 3D model's surface,and coarsely register the 3D model.Then we add the iteration factor and the rotation angle to improve the ICP algorithm to finish the fine registration.Experiments show that the accuracy of the registration algorithm is improved,the convergence speed of the iteration is faster,and the whole process needs no human intervention.