Combining Artificial Intelligence Algorithms to Optimize the Digital Innovation Design of Dong Brocade Patterns Research

AI technology in the development and application of traditional texture recovery and reproduction, deep learning models for traditional texture information and color information consistency migration is still deficient, this paper by using the visual Transformer network advantage and visual Transformer network Transformer encoder structure optimization.That is to say, in the Transformer encoder, the multi-head self-attention module and feed-forward network module are called to process the input data and extract the image features, and then join the edge preservation smoothing technology to remove the strong edge information, preserve some weak edges and local colors, and generate the image texture information with the input texture.The color interpolation method is used to achieve the consistency of texture color texture and image texture migration.The result images of Dong brocade texture style migration show that the image texture migration model based on visual transformer is more capable of generating images with the best style loss value and the best content loss value, and is able to obtain more than 70% of user preference.