Modelling urban environments with voxels: opportunities and challenges

Professor Sisi Zlatanova

University of NSW, Sydney


Contemporary digital 3D city models produced by companies, governmental agencies commonly represent man-made structures using vector representations and do not provide means to analyse the continuous space between them. The existing analytics are usually limited to two dimensions (2D), accompanied by human interpretation. Computational limitations of the traditional vector 3D representation and corresponding data structure are often regarded as the key for the limitations. An alternative to 3D vector representation is 3D raster (voxel) representation. Voxels are the volumetric equivalent of pixels, forming regularly spaced sample sets in 3D space. The merit of such 3D raster is the unification. Objects, spaces or point measurements are represented by one primitive type (voxel), instead of by compounds of different geometric types. Voxels has proven to facilitate spatial analytics: substances of the same type or concentration are easy to identify and group, relationships are strictly defined, interpolations and propagation algorithms can be standardised, operations such as volume or density computation become a matter of counting voxels.

Can this approach be applied to urban environments? Voxel models have been commonly applied to represent continues phenomena like in medical or geological applications. Voxels are increasingly used for game-like, ‘first person’ building of cities such as Minecraft. Voxels have been used to facilitate processing of point clouds, specifically for indoor modelling. Discrete global grids, which are analogous to voxel environments, are being developed for localisation and identification nationally and across the globe. However, voxels have not been largely utilised yet for modelling of urban objects and the spaces between them. Although the advantages of voxels are rapidly becoming acknowledged, the common objection is that the data volume of the required number of voxels hinder the representation of geographic scenes. Moreover processing the enormous data volumes, as a result of voxelisation of an entire city, is computationally expensive.

This talk will address the above mentioned issues, demonstrate applications with voxelised urban scenes and discuss possible opportunities and challenges.


Zlatanova is a SHARP Professor and Head of the Geospatial Research Innovation and Development (GRID) lab at School of Built Environment, Faculty of Arts, Design and Architecture, UNSW. She is internationally recognised expert in three-dimensional (3D) geospatial data modelling, 3D topology, spatially enabled Digital Twins, 3D data structures and specifically frameworks for 3D indoor modelling and navigation. The GRID lab aims to further advance the theoretical foundations of 3D geospatial data modelling and analysis. With a team of talented researchers, she investigates alternative approaches for digital representation of urban environments to ensure intelligent data management and meticulous analysis of the comprehensive interrelations between physical environment, dynamically changing phenomena, and human wellbeing. The GRID team is involved in agile and collaborative research & development, which makes 3D geospatial data FAIR (Findable, Accessible, Interoperable and Reusable)


Jul 07 2021


11:00 am - 12:00 pm

Local Time

  • Timezone: America/New_York
  • Date: Jul 06 2021
  • Time: 9:00 pm - 10:00 pm