Unmanned Aerial Vehicle (UAV) technology has emerged as a transformative tool for 3D reconstruction, offering diverse applications in urban planning, heritage studies, infrastructure monitoring, and emergency response. Despite considerable progress, the field of heritage studies faces challenges due to the scarcity of realworld data tailored for heritage preservation. To address this gap, this thesis presents Heritage3DMtl, an extensive multi-modal dataset comprising 17 heritage buildings in Montreal, acquired using a UAV. The dataset includes images, estimated camera poses, and reconstructed 3D data (point clouds and meshes), providing great detail and diversity.
A Standard Operating Procedure (SOP) for data collection is provided, demonstrating the efficient use of low-cost consumer-grade UAVs to capture heritage buildings. This SOP serves as a replicable blueprint for future similar efforts. Various 3D reconstruction techniques are explored and experimented with using the dataset. Additionally, the dataset’s applicability is showcased through the reconstruction of Level of Detail (LOD) models in alignment with the CityGML standard.
Furthermore, the integration of advanced reconstruction techniques, such as NeRF and Gaussian Splatting, has revolutionized the way we visualize and interact with building sites in digital environments. These techniques enable the generation of photorealistic renderings as well as interactive 3D models, enhancing our ability to study and interpret heritage buildings with unprecedented fidelity and detail.
In summary, this work contributes to the discourse on 3D heritage reconstruction by introducing an open-source dataset that enhances resources available to researchers and practitioners. Heritage3DMtl facilitates advancements in the field and serves as a valuable asset for digital preservation efforts, providing a comprehensive foundation for future research and innovation in heritage documentation and 3D reconstruction.