【作者】 曹房轩； 刘启波；

【Author】 Cao Fangxuan;Liu Qibo;School of Architecture,Chang’an University;

【机构】 长安大学建筑学院；

【摘要】 本研究针对汉唐丝绸之路文化遗产数字化复原中史料匮乏、细节重构困难等问题,提出BIM与生成式AI协同的创新框架。研究以新疆克孜尔尕哈烽燧遗址为对象,通过LOD 3.0级BIM模型构建毫米级精度空间语义底座,结合轻量化SD-Lo RA模型对小样本历史图像(敦煌壁画、唐代墓葬纹饰)进行风格迁移训练,实现遗址"形-纹-境"一体化动态复原。结果表明,该方法可显著提升复原效率,有效解决史料缺失导致的不确定性,为线性文化遗产数字化保护提供标准化技术路径,支持国家"数字丝绸之路"战略。更多还原

【Abstract】 This study proposes an innovative framework integrating BIM and generative AI to address challenges in digital restoration of Han-Tang Silk Road cultural heritage,such as scarce historical data and intricate detail reconstruction.Focusing on the Kizilgaha Beacon Tower site in Xinjiang,a millimeter-precision spatial semantic base was constructed using an LOD 3.0 BIM model.A lightweight SD-Lo RA model was fine-tuned with limited historical images(e.g.,Dunhuang murals,Tang tomb paintings)for style transfer,achieving dynamic "form-texture-context" integrated restoration.Experiments demonstrate significantly improved texture matching and efficiency,effectively mitigating uncertainties from data gaps.This approach offers a standardized workflow for linear cultural heritage conservation,supporting China’s "Digital Silk Road" strategy.更多还原