Aiming at the problems of insufficient pattern restoration, lack of process semantics and weak interactive verification in the digital modeling of Jilin traditional craft cultural heritage, a digital twin system for aesthetic education practice in colleges and universities was constructed. The system collects object images, point clouds, production videos, process texts and interaction logs, and completes multi-source data coding, image point cloud registration, dense reconstruction, mesh optimization, texture mapping, action semantic coding and state synchronization. The Jilin-CraftDT dataset is constructed, which contains 50 craft objects, 2500 multi-view images, 50 sets of point clouds, 800 local images of patterns, 120 production videos, and 4800 interaction logs. Comparative experiments show that the average reconstruction error of the proposed system is reduced to 1.16 mm, and the texture similarity is 94.0%. The frame rate of the five technology scenarios was maintained at 58.4-66.7fps, and the video memory occupation was 3.1-4.6GB. The interactive task completion rate reaches 91.0%, the average response delay is 91 ms, and the state synchronization accuracy is 95.6%. The results show that the system can provide modelable, interactive and verifiable technical support for the revitalization of Jilin traditional craft cultural heritage and aesthetic education practice in colleges and universities.
