Research On Image Classification And Retrieval Technology Of Terra-Cotta Warriors Fragments Based On Multi-feature

With the rapid development of the Internet industry,information and resource sharing have become more and more popular,and information technology has become more diverse and mature.How to protect and display the cultural heritage under the information age has also become a hot topic for major museums.As one of China's large-scale museums,the Terra-Cotta Warriors Museum has the world-famous terracotta warriors.How to protect and display these cultural relics in a digital way,combined with advanced science and technology,is an urgent problem for the Terra-Cotta Warriors Museum of Qin Shihuang.It is one of the measures to protect the cultural relics of Terra-Cotta Warriors by using the computer aided splicing and repairing its fragments.Classifying Terra-Cotta Warriors fragments according to their body parts is an important step in splicing and repairing debris,which can effectively reduce the search space of automatic stitching algorithm and improve the accuracy of stitching.Due to the large workload of artificial fragmentation classification,computer aided automatic classification of cultural relics can reduce the heavy workload generated by manual classification.There are many methods for computer-assisted fragment classification,but these methods are complicated to operate and the readiness for classification is not high.Aiming at the above problems,a classification method of cultural relics fragments based on multi-feature and support vector machine(SVM)is proposed in this paper.In addition,in order to display and disseminate Terra-Cotta Warriors on the Internet,and avoid search engine retrieval of irrelevant resources,museums need to build their own cultural relics digital platform.The main contents are as follows:1.A method of combining the SIFT features with Hu invariant moments for the image representation of the Terra-Cotta Warriors fragments is proposed.On this basis,a multi-feature dynamic weighted fusion algorithm is proposed,which can represent the characteristics of specific body parts of the Terra-Cotta Warriors fragments fully,carefully and uniquely.2.A classification method of Terra-Cotta warriors based on multi-feature fusion and Support Vector Machine(SVM)is proposed,which can obtain the optimal classification model.The input is the feature after fusion and can uniquely represent the fractal characteristics of specific body parts of Terra-Cotta Warriors.3.The image retrieval experiment of Terra-Cotta Warriors fragments has been completed by combining the similarity matching algorithm and multi-feature.Using different similarity measurement techniques and matching algorithms to perform comparative experiments on the image of the Terra-Cotta Warriors fragments.The results show that it can improve the accuracy of image matching and obtain accurate search results by combining classification learning methods and its many features.4.A classification system of the the Terra-Cotta Warriors fragments has been designed and implemented.The classification method has high classification efficiency and accuracy by using the classification model to identify automatically the category of the corresponding Terra-Cotta Warriors fragments.