Research on Virtual Costume Design for Traditional Cultural Symbols: An Intangible Heritage Display Method Based on Computer Vision Technology

Abstract

Intangible heritage carries the historical memory and cultural genes of human beings, and the rise of computer vision technology brings new opportunities for its protection and inheritance. In this paper, we start from the application of computer vision technology to the display of intangible heritage, and introduce the convolutional neural network VGG-19 optimization model into virtual clothing design to construct a clothing image style migration model based on traditional cultural symbols. Experiments show that the model's index results in peak sex-to-noise ratio, structural similarity, root-mean-square error and style migration speed are better than the performance of the comparison methods, and its peak sex-to-noise ratio is improved by 5.09%~51.43%, which proves the high quality and high integration effect of the style migration images generated by the model. When using images containing different traditional cultural symbols for apparel image generation, the non-heritage cultural images with better migration effects in terms of texture color, artistic rendering effect, and creativity of the apparel are the celadon style, the New Year's painting style, and the Dunhuang mural style, with the support rate of the respondents being 49%, 35%, and 39%, respectively. The proposed series of methods can assist the design of clothing styles and broaden the application scope of non-heritage traditional cultural symbols.