Construction And System Implementation Of Three Traditional Dragon Pattern Dataset

The fine traditional Chinese culture is the concentrated expression of the Chinese people’s ideas,customs,lifestyles and emotional styles.The traditional dragon culture plays an important role in the five thousand years of cultural inheritance in China.In this paper,the Internet is used as the research medium to construct the network dataset,which takes the Chinese dragon pattern as the core.The main research contents of this paper include:(1)Dragon pattern dataset construction.Firstly,the storage model of the Chinese cultural heritage specimen bank was constructed as the standard for the entry of images into the specimen bank.The candidate dragon print images were obtained by using web crawler technology.The pre-background of images was separated to eliminate the influence of background color differences caused by different collection standards.Secondly,SVM classifier was used to filter image data.A new texture extraction method with spatial and color information(WSACI-LBP)was proposed,and the features were fused with image invariant distance with the classification accuracy to form a new feature fusion framework.Finally,the perceptual hash algorithm is used to de-duplicate the image pairs in the dragon pattern dataset with the relationships of scaling,translation and rotation.(2)Dragon pattern dataset optimization.The data set of dragon pattern was optimized from two aspects of image label information and image resolution.In the storage model of the specimen library,the image label must be complete.Therefore,image multi-label labeling algorithm based on graph convolution network is used to label the images in the dataset.Then,attention mechanism is added to the existing residual-dense network(RDN)to make better use of high-frequency information in the image,so as to improve the effect of the network on image super-resolution reconstruction.The super-resolution algorithm of residual-dense network(AM-RDN)based on attentional mechanism proposed in this paper is compared with the original RDN on the data set of this paper to prove the effectiveness of the proposed super-resolution reconstruction network of AM-RDN.(3)The dragon pattern dataset system is designed and implemented.The system will be based on the web crawler image acquisition algorithm,image preprocessing algorithm,image data filtering algorithm,image data deduplication algorithm,image data multi-label labeling algorithm and image super-resolution reconstruction algorithm to constitute the three traditional dragon pattern dataset generation process.This paper establishes a set of construction system from image data acquisition,image data filtering,image data deduplication to image data optimization based on the construction of three traditional dragon pattern data set,providing methods and ideas for the construction of Chinese cultural heritage specimen bank.