3D Virtual Reconstruction Of Cultural Relics Based On The Inner Structure And Its Application

The automatic detection of symmetry,regularity and repeatability in 3D models is the fundamental issue in the field of shape analysis and pattern recognition,and it has widely applied in computer graphics.As typical man-made objects,cultural relics contain the genes and spectrums of human civilization,with complex geometric shapes and diverse structural elements.After years of natural erosions and human destructions,many cultural relics have lost their original shapes and integrities.Therefore,the technology of automatically identifying and extracting the internal structure of the 3D models,as well as its applications in virtual reconstruction of damaged cultural relics,has important theoretical and practical values.In this dissertation,the fragments of damaged cultural relics with complex geometric shapes are taken as the research object;the feature recognition of geometric and texture intrinsic structures of 3D models is analyzed;and the method of virtually reconstructing damaged cultural relics is proposed based on the intrinsic structure of 3D models.The main research work of this dissertation is as follows.1.To automatically identify the internal geometric features of the 3D point cloud patterns,the method of extracting the main curvature variation features of regular point cloud is put forward in this dissertation.Firstly,analyze the the main curvature variation law of discrete surface to obtain the main curvature variation characteristics and similarity index of regular point cloud,and then construct the frame to extract the internal geometric structure of the 3D point cloud patterns.This method can effectively extract the geometric features of different points of cloud elements.2.To accurately register the iterations of the muti-resolution points cloud sequence,the method of accurately registering the 3D points cloud is put forward in this dissertation.First,basing on the internal geometric features of the 3D point cloud patterns and Mahalanobis distance calculations,search for the nearest point;and then accelerate the speed of back projection;finally,make corrections by using iinterpolation method.The method proposed in this dissertaiton can avoid the incorrect registrations due to different resolutions,and meanwhile effectively restrain the accumulative errors of sequence iterative registrations.3.To effectively extract the feature lines of the surface with much noise,the method of line drawing based on visual curvature estimation of the cultural relics is proposed in this dissertation.First,estimate the discrete curvature on triangular mesh vertex according to the numbers of etrememum points of height function;and then accurately extract the featured areas of the internal geometric structure of the 3D models.This method can avoid the uneven drawing lines and improve the effect of the detailed line drawing of cultural relics.4.To correctly match the numerous broken fragments of cultural relics from archaeological excavations,the method of automatic classification of broken fragments is put forward in this dissertation.Based on the mixed feature vector and taking the multi-scale integral invriants as the feature representation methods for the contour curves and fracture surfaces,the initial matching point pairs are obtained by using the consistent constraint method.Then,the filtering algorithm is used to remove the pseudo matching points,and the geometric hashing algorithm is used to search the most suitable matching.This method can improve the accuracy and efficiency of matching the broken fragments of cultural relics.