摘要
中国古陶瓷工艺精湛、种类丰富,具有独特的艺术魅力和文化内涵。因此,古陶瓷的科学鉴定一直是文物鉴定研究的热点。针对当前古陶瓷断代方法的不足,在人工智能的大背景下,提出利用深度学习的卷积神经网络(CNN)对古陶瓷文物图像进行器型分类和断代的方法。该方法打破了"古陶瓷断代特征量确定依赖人工完成"的技术瓶颈。测试试验结果表明:该方法对古陶瓷器型分类和断代的准确率达到96.37%,可以作为古陶瓷鉴定的有效辅助手段。
Chinese ancient ceramics are characterized by exquisite craftsmanship,rich variety,unique artistic charm and cultural connotation.Therefore,scientific identification of ancient ceramics has always been the focus of cultural relic identificat ion research.Noting the shortcomings of the current dating methods for ancient ceramics,in the context of artificial intelligence,a novel approach to classify and identify the dating of the ancient ceramic artifacts by using convolution neural network of de ep learning was proposed.The method breaks the technical bottleneck that"the characteristic quantity of ancient ceramics depends on manual determination".Experiment results demonstrate that the accuracy of this method for classification and dating of ancient ceramics is 96.37%,making it an effective auxiliary method for dating ancient ceramics.
作者
冯金牛
周强
张瑞瑞
王莹
罗宏杰
FENG Jinniu;ZHOU Qiang;ZHANG Ruirui;WANG Ying;LUO Hongjie(School of Electrical and Control Engineering,Shaanxi University of Science&Technology,Xi'an 710021,Shaanxi,China;Research Institute of Silicate Cultural Heritage,Shaanxi University of Science&Technology,Xi'an 710021,Shaanxi,China;School of Material Science and Engineering,Shanghai University,Shanghai 200444,China)
出处
《陶瓷学报》
CAS
北大核心
2022年第1期145-152,共8页
Journal of Ceramics
基金
国家重点研发计划(2019YFC1520100)
陕西省科技计划项目(2019GY-090)。
关键词
古陶瓷断代鉴定
卷积神经网络
深度学习
chronological identification of ancient ceramics
convolutional neural network(CNN)
deep learning
