Research And Implementation Of Digital Reconstruction Method For Traditional Pattern

Traditional pattern is the crystallization of human culture,is an important part of the culture of a nation,a country.How to provide material for modern design based on the traditional pattern,let tradition and fashion combine,to carry forward and protect the traditional culture is an urgent problem to be solved at present.Based on the traditional pattern,this paper proposes a rule-based geometric pattern generation algorithm and a semantic-based traditional pattern generation algorithm,which not only preserves the characteristics of the traditional pattern,but also realizes the diversified generation of the traditional pattern.The research and development of the traditional pattern generation system can also easily and quickly generate the pattern materials for the Chinese culture material library continuously.The research content and innovation points mainly include:(1)Generate traditional geometric patterns with simpler structure,strong symmetry and strong regularity.From the perspective of mathematical geometry,this article deconstructs the basic elements of the pattern,analyzes the pattern's generation rules,summarizes the corresponding mathematical formulas,and realizes the parameterized generation algorithm.By adjusting the optional input parameters,it is possible to generate similar patterns with similar styles but with different attributes.In the generation of geometric pattern rules,a traditional Chinese pattern---Turtle pattern,and a typical foreign pattern---Kolam pattern were selected.Using mathematical geometric analysis and parameterization,the generation algorithm of turtle back pattern is proposed;with the help of predecessor's classification method,the method of generating Kolam pattern is realized by using matrix rules.(2)For other types of patterns with complex composition and difficult to find the generation rule,it is not suitable to use the same method to analyze and summarize as the generation of geometric patterns.In order to solve this problem,the generative adversarial network model,which is popular in recent years,is used in this paper to complete the generation of patterns,based on the basic network structure of Pix2PixHD.Combining self-attention with spatial adaptive normalization layer in SPADE,a new residual block structure is proposed to realize the redistribution of feature weights and avoid the semantic loss in the process of normalization.On the other hand,in order to improve the distortion of the internal structure of the generated image,this paper adds edge-consistency loss,which is used to calculate the edge loss between the generated image and the original image to guide the generation of the image.Experimental results show that the proposed method can improve the quality of image generation to a certain extent.This paper also makes a comprehensive comparison on the objective evaluation indexes such as ACCU,FID and MIOU.(3)Designed and implemented a traditional pattern generation system,with rule generation and semantic generation algorithms as the core.The system can support the rapid generation of traditional geometric patterns of the same type in a parameterized manner,and can also generate traditional pattern images by drawing semantic masks,providing users with the basis for material creation.Provides the function of online saving and downloading of pictures.After completing the user mapping association with the WeChat applet,the secondary creation of the material can be realized in the associated WeChat applet.