Digital Heritage Conservation for Historical Villages: UAV–LiDAR Fusion and AI-Based CV Material Analytics
Digital Heritage Conservation for Historical Villages: UAV–LiDAR Fusion and AI-Based CV Material Analytics
Speaker:
Junpeng Fan
Summary:
This study presents a digital heritage framework for endangered Chinese villages. Combined with UAV photogrammetry, LiDAR point clouds, and a Rhino-Grasshopper fusion workflow, the system generates coherent multi-scale 3D models that overcome shadow and terrain discontinuities. Using an advanced computer vision deep-learning model-SAM2, we achieve automated semantic segmentation and material recognition, enabling precise and non-invasive facade analysis. A Unity-based platform enhances visualization and public accessibility. The framework reduces manual processing and hardware demands, offering a scalable, high-precision solution with significant value for long-term digital conservation and future AI-driven reconstruction
This talk was delivered online on 16 November 2025. Talk duration is 1 hour and 17 minutes.
Speaker Bio:
Junpeng Fan holds a Master’s degree in Architecture from Politecnico di Milano and is currently a PhD candidate at Keio University, jointly affiliated with the University of Tokyo’s Ikeda Lab and Keio’s Kobayashi Lab. His research focuses on applying digital technologies and CV-based AI to the conservation of Chinese heritage villages.
CPD Details:
This talk is offered as a formal CPD activity for architects, mapped to performance criteria PC18 and PC28 in the Australian 2021 National Standard of Competencies for Architects.
A copy of the CPD questions can be accessed here. If you take notes in digital format during the video, you will be able to copy/paste your answers into the questionnaire at the end.
Upon completion of the video, a link to the CPD questionnaire will be unlocked. Submit your answers via that link to earn 1 formal CPD point and receive your certificate for record-keeping purposes. Note: The questionnaire link will only load if the video is viewed from this webpage. If you navigate to YouTube to view the video, you will not be able to access it.