Research And Application Of Terracotta Warriors Registration Based On 3D Point Cloud

In the process of 3D reconstruction of the Terracotta Warriors models to establish a complete simulation model,computer-aided virtual restoration,etc.,due to the characteristics of the Terracotta Warriors and the limitations of the 3D laser scanner,it is necessary to use multiple sites to obtain point clouds at different coordinate angles.The point cloud models are registered to form a complete simulation model of Terracotta Warriors.The result of registration will directly affect the effect of 3D reconstruction and subsequent applications.This thesis studies the registration of Terracotta Warriors based on 3D point cloud.The main research contents are as follows:(1)Aiming at the problem of excessively large point cloud and time-consuming search for registration point pairs,a point cloud initial registration algorithm based on intrinsic shape signature and feature histogram is proposed,and the key point detection and multiple feature descriptors are combined differently.Establish a similarity measurement function to determine the initial registration point pair.Further,an improved SAC-IA is used to complete the initial registration.Experiments show that the proposed algorithm has a good effect on the initial registration process of point cloud data from different angles.It reduces the search space for searching for registration point pairs,correspondingly reduces the time for searching for registration point pairs,and improves the initial registration rate.(2)Aiming at the problem of iterative closest point algorithm that is easy to fall into local optimal and large angle deviations in the initial registration,this thesis proposes the fusion angle and distance constraint FADC method.In the precise registration stage,the distance and angle constraints are introduced to perform the second registration on the pairs filter.Set different thresholds to eliminate point pairs that do not meet the FADC constraints.At the same time,Kd Tree search strategy is adopted to search for neighboring points.Experiments show that this algorithm reduces the point-to-point search space,improves the speed of searching for neighboring points,and performs secondary removing on the registration point pairs,which further improves the registration accuracy.(3)Design and implement the Terra-Cotta Warriors Point Cloud Oriented Registration System.The system realizes the functions of Terracotta Warriors point cloud preprocessing,key points extraction,initial registration and precise registration.Experiments show that the overall function of the system is perfect,and the key points of large-scale point cloud data can be extracted and registered under the condition that the structure of the point cloud model is basically unchanged.It can be practically applied to 3D reconstruction of Terracotta Warriors.The research in this thesis is supported by the key industrial chain project of Shaanxi Province:Digital Museum Key Technology Research and Development and Terracotta Warriors Wisdom Museum Prototype Demonstration(2019ZDLSF07-02).