Research And Implementation Of Cultural Relics’ Recognition System Based On Convolutional Neural Network

The Chinese civilization has a long history.It has formed a splendid culture for thousands of years,and has left countless historical relics.They not only condense national pride,but also are precious materials that study the lifestyle of ancient people.Therefore,they are identified And protection is particularly important.In recent years,with the advancement of science and technology,image-based cultural relic recognition technology has also been developed to some extent.The collection and annotation of cultural relics data sets have always been an important part in the research of cultural relics identification,but because there is no unified cultural relic image data sharing platform,each research team may repeatedly collect and label the same cultural relics data,resulting in a waste of human and material resources.Based on the above background,this paper proposes an improved convolutional neural network structure for cultural relics recognition,and designs and implements a cultural relics recognition system.On the one hand,it provides online accurate and rapid identification of cultural relics,on the other hand,it provides a cultural relics image data sharing platform to cultural relics researchers.The main research work of this paper is as follows:1)Acquire cultural relic image data including 4 categories of ancient currency,porcelain,jade,bronze,etc.through reptiles and other technologies,and manually filter to obtain the original image data set of cultural relics.2)Use two methods to preprocess the original data set.One is to perform image segmentation on the original data set,then normalize the image size,and obtain the W21 data set through data enhancement operations such as translation and flipping.The second is to directly normalize the image size of the original data set and perform data enhancement flipping at different angles to obtain W22.3)Based on the classic networks VGGNet and ResNet,an improved convolutional neural network structure is proposed.The network has 20 convolution kernels and a total of 9 convolutional layers.Through training,the improved convolutional neural network structure can achieve 98.537%and 97.830%accuracy on the W21 and W22 data sets,respectively,which is significantly improved compared to the VGGNet 16,ResNet34 and other networks used in the comparative test in this paper.4)Design and develop a set of cultural relics recognition system based on convolutional neural network.The cultural relics identification system is mainly divided into two parts:a cultural relics identification website and a cultural relics identification management system.The cultural relics identification website can provide online identification capabilities for cultural relics images.The cultural relics identification management system mainly provides cultural relics identification researchers with a shared cultural relics identification image data and technology Platform.The front end of the system is developed based on the current mainstream framework React,which can realize the sharing of cultural relics data sets and the online recognition of cultural relics images.