In this paper, the semantic description framework is used to standardize the extraction of semantic information of non-legacy images. The SIFT algorithm is chosen to calculate the key feature points of non-legacy images. The integrated semantic description framework and SIFT algorithm construct a model to extract the non-heritage image features globally and process them locally, and add the attention feature fusion module to fuse the features that are inconsistent in semantics and scale, so as to realize the accurate extraction of features. Use the algorithmic model of this paper to extract the color features of She ribbons. Develop a website for she color band design and verify its usability. Collect website evaluation data from target users to study the role of digital translation of non-heritage elements. The color feature extraction results are richest and most detailed when the number of She ribbon feature colors extracted is 21. The website usability scale score was 50.31, rating B+, with usability. 65% of the users thought that the website embodied the cultural characteristics of She ribbons. 71% of the users thought that the website was very helpful for understanding the ethnic graphic culture. 88.16% of the users thought that the digital design of She ribbons could effectively promote the dissemination of the ethnic graphic culture.