Research On Cultural Relic Recognition Based On Feature Pyramid And Attention Model

Cultural relics are the epitome of a country’s history and culture,each country and even the entire human civilization in the history of non-renewable precious resources and historical heritage.With the development of archaeology,more and more archaeological relics have been excavated,which has brought great challenges to the classification and sorting of archaeological relics.At the beginning of our social and economic development,we will continue to send a large number of personnel,spending a lot of materials and money to protect cultural relics.Our country’s cultural relic protection work has obtained the certain result at present,but the existence question also has many.The general public and even a few museum staff know little about the cultural relics,which is seriously adverse to the protection of cultural relics in our country.All walks of life are in-depth study on how to carry out more comprehensive protection of cultural relics.The development of identification and classification technology of cultural relics is helpful to better identify and classify cultural relics,let more people and staff know about them,prevent them from losing,and promote the protection of cultural relics.The main contents of this paper are as follows:(1)In view of the characteristics of the dense spatial distribution and small targets of cultural relics,the target detection method is applied to the task of cultural relics target detection.Based on the feature pyramid structure,Dense Net is selected as the backbone network,the target detection network and the corresponding training strategy are studied,and a new feature fusion module is added.(2)In order to solve the problems of insufficient extraction of feature information and low robustness of feature similarity measurement for cultural relics in current cultural relics identification methods,the Res Net network is selected as the backbone network,and the self-attention mechanism based network for extracting the apparent feature of cultural relics is studied,so that the network can focus on key areas in cultural relics images and extract the apparent feature of cultural relics for identification and classification.Finally,experiments are designed to analyze and verify the methods of detection,extraction and classification of cultural objects.It proves the validity of the feature pyramid based method and the self-attention based method,and evaluates the performance and effect of each part.Through the comparison with the related algorithms,the advantages and disadvantages of the method are analyzed.