Geomatics Day - Student project 4
Semantic 3D Point Cloud Generation for Heritage Buildings Using Gaussian Splatting
This project will focus on the generation and visualisation of 3D point clouds of heritage buildings utilising Gaussian Splatting to create semantically labelled smart point clouds. By capturing and processing highly detailed 3D spatial data, the project will contribute to the digital preservation, analysis and interpretation of heritage structures.
Gaussian Splatting is a technique that enhances point cloud representation by converting points to continuous visually smooth Gaussian functions using Gaussian kernels to allow for a better representation of real world scenarios by improving the rendering quality and visual coherence in the point cloud data.
The project will aim to explore various existing Gaussian Splatting tools and evaluate their effectiveness in representing complex structural components of buildings. Through the application of semantic labelling with machine learning, this project aims to aid in the usability of point clouds as BIM models and increase the ability to provide contextual information regarding specific building features without the need for manual labelling.
