基于深度学习的丝绸文物纹样识别应用  

Application of silk cultural relic pattern recognition based on deep learning

作　　者：孙选铭 苏淼[1,2] SUN Xuanming;SU Miao(College of Textile Science and Engineering(International Institute of Silk),Zhejiang Sci-Tech University,Hangzhou 310018,China;World Silk and Silk Road International Research Center,Zhejiang Sci-Tech University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学纺织科学与工程学院(国际丝绸学院),杭州310018 [2]浙江理工大学国际丝绸与丝绸之路研究中心,杭州310018

出　　处：《丝绸》2023年第8期1-10,共10页Journal of Silk

基　　金：国家重点研发计划课题项目(2019YFC1521301);浙江省文物保护专项项目(2021016);浙江理工大学科研启动基金项目(20202214-Y)。

摘　　要：为对数字化文物进行快速自动分类,提高藏品数字化进程,加快数字博物馆的构建,文章利用深度学习对丝绸文物的纹样进行自动识别。依据实物纹样的分类方法,建立了包含花卉纹、飞鸟纹、“卐”字纹、云纹四类纹样的样本库。利用VGGNet、ResNet、MobileNet实现对纹样的分类,结合Faster R-CNN、YOLOv5、SSD目标检测算法实现对纹样的识别与定位。实验结果表明,MobileNet对丝绸纹样分类的mAP达到83.51%;在目标检测算法中YOLOv5的识别与定位效果最好,其mAP为88.42%。与通过人工分类相比,采用深度学习算法进行分类与识别,可以在降低难度的同时提高分类的速度和准确率,为纺织品文物的鉴定与保护提供了新的思路。Silk,as one of the greatest inventions in ancient China,has a history of over 5000 years and represents one of China’s most significant traditional cultures.The value of ancient silk is a historical witness,and the important information it carries is also important.As the most representative cultural symbol of a nation,decorative patterns are a crucial part of the information carried by ancient silk.Therefore,extracting information from ancient silk is of great significance for the protection and promotion of China’s excellent traditional culture.Although museum visits are the most effective way for non-cultural relic protection workers to access ancient silk,this approach is not only inconvenient but also insecure,particularly in the context of the recent worldwide outbreak of the new coronavirus.To break the spatial limitations of collections and museums,bring people closer to cultural relics,realize resource exchange and sharing,and serve the promotion of the Silk Road spirit,China National Silk Museum has issued an initiative to jointly build the Silk Road Online Museum(SROM).The classification of cultural relics is an important step in the construction of digital museums.Compared with ordinary products,the classification of cultural relics is difficult.Specifically,textile cultural relics are more difficult to classify because of the difficulty in preserving materials and poor integrity.Therefore,how to quickly and accurately complete the identification and classification of textile cultural relics has become a problem to be solved.In view of the above problems,we adopted deep learning to automatically identify the patterns of silk cultural relics.According to the classification method of physical patterns,a sample library including flower patterns,bird patterns,swastika patterns and cloud patterns was established.VGGNet,ResNet and MobileNet were used to classify patterns,and Faster R-CNN,YOLOv5 and SSD target detection algorithms were used to identify and locate patterns.The results show that MobileNet

关 键 词：丝绸文物 深度学习 数字博物馆 目标检测 神经网络 自动分类 

分 类 号：TS101.1[轻工技术与工程—纺织工程]