Miao Embroidery Pattern Redesign Method Based On Deep Learnin

Miao embroidery is one of the traditional Chinese handicraft techniques with a long history and rich connotations.It was included in the first batch of national intangible cultural heritage protection list in May 2006.The inclusion of Miao embroidery on the first national list of intangible cultural heritage was determined by its rich cultural connotations,spiritual values and the plight of those who were left behind in the modernisation process.The traditional Miao embroidery skills must be protected and passed on to the next generation,as they are endangered.Similarly,Miao embroidery pieces,which are a product of Miao embroidery techniques,also need to be protected.The existing methods of storing Miao embroidery pieces are mainly hand-painted and photographed,but it is still a challenge to preserve and diversify the design of Miao embroidery pieces.In summary,this paper uses computer technology to collect,classify,image segmentation and migrate styles of Miao embroidery pieces to expand the application of traditional Miao embroidery styles in different fields,and to help the inheritance and digital conservation of Miao embroidery,with the following research and results:(1)Taking Miao embroidery pieces from Qiandongnan as an example,the artistic characteristics of Miao embroidery in Qiandongnan are analysed from the perspectives of cultural connotation,combination elements and regional characteristics,while Miao embroidery in Qiandongnan is summarised and classified according to metadata standards.The collection of a large number of embroidery patterns through fieldwork and photography provides data support for the subsequent computer technology in image processing,as well as a standardised and reasonable database for the collection of Miao embroidery patterns;(2)The embroidery patterns of Miao embroidery pieces obtained by photographic techniques are often blurred,reflective,wrinkled,and indistinguishable between foreground and background,which poses a challenge to the digital storage of Miao embroidery.In view of these storage problems,this study combines the semantic segmentation of images in computer technology with traditional Miao embroidery patterns,and proposes a multi-scale semantic segmentation architecture that can effectively segment the texture,lines and outline of embroidery patterns,so as to quickly determine the main contents of the embroidery pieces and provide more means for subsequent processing,fusion and redesign;(3)Based on the existing Miao embroidery database,the project aims to expand the colour characteristics of embroidery patterns through computer vision techniques,and to combine traditional Miao embroidery patterns with Dunhuang caisson and Western art.A network combining Convolutional Neural Networks(CNN)and Transformer in Natural Language Processing(NLP)is proposed to effectively extract image features through a parallel architecture,laying the foundation for subsequent image style transfer.The results of this study show that the proposed network is based on the CNN and the Transformer in NLP.The research results show that the proposed multi-scale feature learning-based semantic segmentation method and the parallel architecture style transfer method have a progressive relationship,which can effectively process,extract information,fuse features and migrate traditional Miao embroidery patterns,and generate patterns that are different from the traditional Miao embroidery style but retain their content characteristics,providing an effective way to expand Miao embroidery techniques and digitise and preserve embroidery patterns.This will provide an effective method for the expansion of Miao embroidery techniques and the digital conservation of embroidery patterns.