nD-PointCloud as novel representation for spatio-temporal phenomena involving massive data sets

Peter van Oosterom (P.J.M.vanOosterom@tudelft.nl)

I propose a new field of spatial information science allowing for breakthroughs in spatial analysis, studies on water management, land use, urban planning, transportation and mobility, human and social geography, and other fields where spatial data is used. Furthermore, I propose a novel nD-PointCloud model for handling massive multi-dimensional (nD) point cloud data sets, representing space, time and added information (colour, material properties, velocity etc.). Current spatio-temporal representations are based on either gridded or object models. Typically, these are organized in a fixed number of levels of importance (detail/scale), introducing serious limitations: fixed level choices and data density jumps between levels. Instead, my approach facilitates continuous levels of importance, which can be regarded as an added dimension to space and time and has the following original and novel aspects:
  1. supporting organizing AND property dimensions;
  2. providing attribute granularity: from individual point to group level;
  3. obtaining the continuous level of importance value for a point;
  4. representing and assessing effect of different coordinate reference systems: spherical or Cartesiann;
  5. enabling smart point clouds.
Using high-resolution nD space filling curves I will realize deep integration of space, time and importance as basis for data organization and apply High Performance/Throughput Computing for big data (trillions (1012) of points). By enabling operations directly on the raw point cloud data, nD-PointCloud largely avoids the extract-transform-load hurdle, which is an increasingly serious problem in using big data. This enables major advances in domains requiring lossless spatio-temporal data of extremely high accuracy, such as applications of: scanned terrain surface/object models and moving object (trajectory) models, which are used as Proof-of-Principle.

The Massive Point Clouds for eSciences project was the predecessor of the current project.

Consortium meetings can be found here.

Additonal information:
  1. Peter van Oosterom, nD-PointClouds: a model for deeply integrating space, time and scale,
    Keynote presentation at the 11th 3D GeoInfo Conference, 20 October 2016, Athens, Greece.
  2. Peter van Oosterom, Martijn Meijers, Edward Verbree, Haicheng Liu and Theo Tijssen, Towards a relational database Space Filling Curve (SFC) interface specification for managing nD-PointClouds,
    Invited presentation at the Munichner GI-Runde, 14-15 March 2019, Munich, Germany.
  3. Peter van Oosterom, From discrete to continuous levels of detail for managing nD-PointClouds,
    Keynote presentation at the ISPRS Geospatial Week 13 June 2019, Enschede, The Netherlands.