Research On Residual Network Identification And Multi-feature Mosaic Technology For Cultural Relics Restoration

The virtual restoration of cultural relics can improve the efficiency of cultural relics restoration and reduce the secondary damage of cultural relics.For the complex cultural relics of the Terracotta Warriors,the identification and matching of the fragments is the key technology.In the identification of debris parts,deep learning has a better recognition effect,but shallow network identification accuracy is low and deep network will have a certain degree of network degradation.In the fragment matching splicing,it is often difficult to achieve better splicing effect based on a single feature for fragments with complex geometric contours.In response to these problems,this paper takes the Terracotta Warriors fragments as the research object and conducts related research.The main research works of this paper are included as follows:(1)In the traditional cultural relics restoration work,the type of debris parts of Terracotta Warriors is difficult to judge.To solve this problem,a three-dimensional residual neural network algorithm for identifying the point cloud type of cultural relics was proposed.In this paper,the three-dimensional convolutional neural network is optimized and improved,and residual learning is introduced to solve the problem of deep three-dimensional convolutional neural network degradation.To some extent,the accuracy of object point cloud type recognition is improved,and the three-dimensional residual neural network is applied.In the terracotta debris point cloud data,the terracotta debris type identification is performed.The experimental results show that the recognition accuracy rate of threedimensional residual neural network method used in the terracotta warrior point cloud type identification task reaches 83.59%,meeting the debris type identification requirements.(2)The artifacts always have complex geometric contour features and using single feature always have low matching accuracy for Terracotta Warriors and Fractures.To solve this problem,a debris splicing algorithm based on contour multi-feature constraints is proposed.The algorithm performs Delanuay triangulation on the scattered point cloud data to get constructed triangular mesh data;Then extracting the three-dimensional contour curve from the constructed triangular mesh data the feature points on the curve;Then,the feature vectors are constructed for different features of the extracted feature points,and a logistic regression method is introduced to fuse the multi-feature construction feature point matching decision function;Then,constructing multi-fragment dynamic matching algorithm by using structural feature point matching decision function and 3D contour forming feature vector;Finally,the three-dimensional contour matching is performed by the debris matching algorithm,and the contour matching fragments are spliced by the rigid body variation and the ICP registration algorithm.Experiments show that the algorithm has good splicing effect for the complete fragmentation of the debris contour,and can complete the splicing work of multiple fragments at one time.(3)The Terracotta Relic Restoration System was designed and implemented.Applying the algorithm to the restoration of Terracotta Warriors and Cultural Relics,the terracotta artifact fragments are recovered from type identification to the overall splicing process,which provides guidance for the splicing of cultural relics.