Research On Automatic Segmentation And Registration Of Fragment Model For Cultural Relics Restoration

Cultural relics,bearing the history of a nation,maintain the cultural identity of the nation.However,with the flourishing development of cultural currently,whether cultural relics can be protected scientifically and reasonably is greatly significant to the inheritance and spread for the history and culture.The restoration of cultural relics is an important part for the scientific and rational protection of cultural relics.The process of cultural relics restoration and its complexity,especially for the joining and reassembling of broken cultural relics fragments,is a very detailed and time-consuming project.This kind of work has a high standard for the restorers of cultural relics and exists many risks.Many people hesitate to be occupied on it,resulting in a serious shortage of restorers.However,through computer technology,the process of cultural relics restoration has been simplified,and manual intervention is also minimized.To achieve automatic reassembly(piecing)for utensil fragments,a local geometric feature learning-based fracture surface extraction and utensil reassembly method is presented in this article whose main substeps are as follows.First,based on obtained 3D fragment models,a triangle cell descriptor is proposed to describe the spatial neighborhood geometric features.Second,Feature Mapping Images(FMIs)are established as a training dataset.Third,after labeling the ground truth data,a convolutional neural network(CNN)is trained using the FMIs.Fourth,based on processing to eliminate mislabeled triangle cells,a skeleton of the fracture surface margin can be generated;Fifth,a shortcut-based strategy is proposed to eliminate residual triangle cells to extract fracture surfaces.Sixth,a control point and vector-based strategy is proposed to complete the matching of fractured surfaces and achieve the prealignment of the fractured surfaces.Finally,a cyclic error iteration strategy is designed to assemble the fragments into a holonomic utensil.This learning-based framework is better at extracting the key geometric data(fractured surfaces)of utensil fragments than are several classical methods.It can not only provide technical support for computer-assisted archaeology but also function as a new method of 3D graph processing.