top of page


3D hyperspectral mapping of complex outcrops and mine faces


Point clouds provide a simple but flexible way of storing, processing and visualising 3D data. Attributes from various sensors (e.g., geophysics, hyperspectral) can be attributed to 3-D points to derive spatially varying scalar fields. However, unlike image data, 3-D point clouds have an ill-defined topology (point cloud can be defined in an unordered set), making the application of many modern deep learning techniques a significant challenge. We aim to mitigate some of these challenges by developing fit-for-purpose deep learning techniques that are compatible with high-dimensional data distributed across arbitrarily complex 3-D point clouds.

bottom of page