Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Fetching contributors…

Cannot retrieve contributors at this time

237 lines (169 sloc) 7.362 kb

Display and Management of Geomatics Research Data

.fx: title

Display and Management of Geomatics Research Data

Timothy Daniel Trewartha and Michiel Johan Baird

oct.png

Zamani Project

  • Started by the UCT Department of Geomatics in 2004
  • Aims to preserve cultural heritage by documenting sites and producing laser scanned models
  • Over 100 models of sites in various African countries including South Africa, Zimbabwe, Kenya and Tanzania
  • Some of the models are very large and contain billions of points
zamani.jpg

Problems Faced

  • Fast-growing volume of data
  • Difficulty in storing the data
  • Difficulty with viewing the large models in real-time
  • Data locality issues
  • Large number of users interact with the data

Solution Approach

  • To enable viewing of the large 3D models at interactive frame rates we will implement a hierachical multi-resolution data structure
  • A server with a large amount of storage will be procured; this will be used to store the models and GIS data, and support the core functionalities of the solution.

Dynamic Viewing of Large 3D Models - Tim

  • The Department of Geomatics has indicated that they have difficulties handling the sizes of some of their models
  • Some of the models they are dealing with contain over 8 billion points
  • At this point, traditional viewing methods cannot cope; the resolution of the original model must be decreased manually beforehand
  • This compromise is often unacceptable
greatzimbabwe.jpg

Research Question

  • Is it feasible to support real time viewing of models containing billions of points?
  • Answering this question in the affirmative would enable exploration of the Zamani models in their full detail
  • It would have a significant impact in the Geomatics department

Proposed Solution

  • Implement a multi-resolution data structure to divide our model into manageable chunks
  • Initially only a subset of the points is available
  • As one zooms into the model and greater detail is required, we dynamically fetch additional points from our data structure until the full original detail is available

Relevant Literature

  • Common approaches to structuring large 3D models include octrees, R-trees, bounding sphere hierachies, and Hilbert Space Filling Curves
Data Structure Largest model rendered Reference
Bounding Sphere Hierachy 8 million points (Rusinkiewicz and Levoy, 2000)
Hierarchy of Tetrahedra 300 million points (Cigoni et al., 2008)
Octree 2.2 billion points (Wand et al., 2007)

Proposed Datastructure

  • From researching the literature it seems that octrees have the best performance
  • All data is stored in the leaf nodes
  • Inner nodes provide simplified multi-resolution representations
  • No leaf node should contain more than a specified number of points
  • Empirically, it seems that a value of around 30,000 gives good performance
oct.png

Evaluation Criteria

  • Can the system render the largest of the Zamani models at interactive frame rates?
  • If this goal is achieved the system will be a success
  • Varying degrees of success can also be determined by testing smaller models of varying sizes

Workbench - Michiel

How effective is an automated workflow solution in GIS context?

workflow.png

Previous work

  • Various fields of science has adopted and implemented workflow systems
  • These systems have increased efficiency and research output
  • GIS research has been shown to be applicable to an automated workflow system; this has not however been implemented

Proposed Solution

  • Use an existing workflow system as various platforms already exist
  • Design a workflow that is applicable to GIS
  • Write middleware to integrate with existing GIS tools
  • Software that automatically transfers data as it is needed down the pipeline

Testing Criteria

  • How much does the content delivery system decrease waiting time?
  • How effective is the workflow system based on the analytics that will be generated be the system.

Division of Work

  • Tim will be implementing the hierachical datastructure to enable real-time interaction with the large 3D models.
  • Michiel will be implementing a Scientific workbench that is specialised for GIS research. This will pay special attention to data movement and intergration with existing GIS tools.

Deliverables

  • GIS workbench
  • Middleware for core functionalities
  • Data Flow Facilitator
  • Hierarchical Data Structure
  • Streaming Infrastucture

Timeline

.fx: timeline

Description Start End
Web Presence 25 May 12 June
Initial Feasibilty Demonstration 11 June 29 June
Background Chapter 2 July 29 July
Design Chapter 29 July 29 August
First Implementation 1 July 29 August
Final Implementation 29 August 28 September
Report Outline Complete 28 September 10 October
Report 28 September 31 October
Poster 31 October 3 November
Presentation 11 November 18 November
Jump to Line
Something went wrong with that request. Please try again.