清末岭南在繁荣的中西贸易和文化交流，以及书院文化、士绅文化等的影响下，形成了以广州为中心、以十三行行商为代表、官商士绅竞相造园的特殊历史时期，其造园艺术达到岭南古典园林发展的巅峰，留下了包括外销画和摄影等在内的大量历史图像。鉴于园林实物遗存较少的现状，这些具有丰富空间信息的历史图像为深入探究清末岭南私园的空间知识提供了可能。目前，依据图像开展园林空间复原的研究尚属起步，利用计算机图形学和人工智能等数字技术对该领域新范式的探索具有重要意义。. 本课题以清末岭南私园为研究对象，兼顾二维历史图像与实物遗存。尝试结合历史研究、实物遗存研究及机器学习算法，在计算机图形学和人工智能技术的支持下对二维历史图像进行三维重建，开展清末岭南私园图像的空间复原研究，并通过实物调查及图像研究获得的空间信息，分析和研究清末岭南私园的空间形态，归纳和总结其空间构成规律，完成其空间知识体系的构建。

In the late Qing Dynasty, as a result of the prosperous trade and cultural exchanges between China and the West, as well as academy culture and gentry culture, Lingnan had a special historical period. In this period, officials and merchants which the Thirteen Factories merchants are the representatives competed to create gardens in the religion with Canton as the center. The garden art reached its summit in the development of Lingnan classical gardens, leaving a large number of historical images including export paintings and photos. Although there are few physical remains, these historical images with rich information provide the possibility to explore the spatial principle of Lingnan private gardens. At present, research on garden space restoration based on images is still in its infancy. It is of great significance to explore new paradigms in this field by using digital technologies such as computer graphics and artificial intelligence. . Referring to 2D historical images and physical remains, the project takes Lingnan private gardens as the object of research. Combining the historical researches, the physical remains researches and the machine learning algorithms, with the support of computer graphics and artificial intelligence the project attempts to reconstruct 3D scenes from 2D historical images on the purpose of spatial restoration research. Based on spatial information obtained from physical investigation and historical images, by analyzing the spatial form and summarizing the spatial principle, the project will be carried out to establish the spatial knowledge system of Lingnan private gardens.