Research On The Classification Algorithm Of Terracotta Warrior Fragments Based On The Optimization Model Of Convolutional Neural Network

With the rapid development of visualization technology and computer digitization,the protection and inheritance of cultural heritage have gradually entered the information age.The terra-cotta warriors and terracotta warriors and terracotta warriors unearthed from qin shihuang's mausoleum and burial pit are numerous in number,complex in geometry and structure and incomplete in surface feature information,resulting in low recovery efficiency.By classifying the fragments of the terracotta warriors and horses and determining the subordination between the fragments,and then splicing within each part to complete the reorganization of the fragments of each part,the problem of high complexity of the virtual splicing algorithm and inaccurate results can be effectively solved.To deal with the above problems,this paper combines the intuitionistic fuzzy set with the improved convolutional neural network model to classify the Terracotta Warriors under the support of the Shaanxi Provincial Natural Science Fund Project—“Research on Adaptive Learning Model of Intuitionistic Fuzzy Group Intelligence for Damaged Terracotta Warriors”.The specific research content is as follows:1.Construct a two-dimensional image data set of Terracotta Warriors debris.A three-dimensional data set is acquired by using a scanning device,and a global multi-view two-dimensional image data set is obtained by using a Geomagic system on the basis of feature annotation;a two-dimensional image data set is expanded by using a Gaussian mixture model conditional generation against network data enhancement algorithm;The expanded data set can greatly improve the issue that the subsequent experimental results are inaccurate and over-fitting due to the limited amount of existing fragment data,and provide strong data support for the subsequent work of this paper.2.The Terracotta Warriors debris dataset was applied to the convolutional neural network model for the first time,and a classification method of terracotta warriors debris based on convolutional neural network was proposed.The preprocessed fragment data set is put into the convolutional neural network for data training,which avoids the time consumption of artificial feature extraction and greatly improves the online calculation efficiency.Experiments show that the method proposed in this paper is more accurate and efficient than the traditional method.3.A classification method of Terracotta Warriors debris based on shallow neural network is proposed.This method takes into account the over-fitting problem caused by the convolutional neural network in the feature extraction of debris due to excessive convolution and excessive extraction characteristics,optimizes the traditional convolutional neural network to a certain extent,improves the feature recognition rate of terracotta warriors debris,reduces the time complexity of the algorithm,and improves the efficiency of fragment classification.4.A classification method based on intuitionistic fuzzy C-means convolutional neural network for Terracotta Warriors is proposed.This method extracts fragment features through shallow convolutional neural network,and then introduces intuitionistic fuzzy sets to better realize the feature separation of terracotta warriors and shards,makes the features more clear,completes the clustering of fragments,and finally uses the prior knowledge to determine the clustered parts.This method reduces the work cost,improves the complexity of the fragment classification work,and improves the classification accuracy.