Feature tracking can be used to correlate extracted objects from
one data set to the next.
What is New
Feature Tracking
on Unstructured Datasets.
FVis5D - Vis5D Enahanced
with Feature Tracking
The evolutionary events can be characterized as follows:
- Continuation
- An object continues in consecutive time steps (with rotation or translation),
as it increase (decreases) in volume or magnitude.
- Creation
- A new object appears.
- Dissipation
- An object disappears.
- Bifurcation
- An object separates into two or more substructures.
- Amalmagation
- Two or more objects merge.
Assumption
Time between successive data sets is small.
Heuristic
Corresponding objects in succesive data sets overlap in space.
Methods
- Correlation of objects is performed based on minimum difference.
- An octree spatial data structure is used to implement the algorithm.
Examples of 3D Volume Tracking
Pseudospectral simulation of coherent turbulent vortex structures with
a 128^3 resolution (100 time steps). Variable being visualized is vorticity
magnitude (thresholded isosurfaces at 48% of maximum).
Simulations by Dr. V. Fernandez
In this video, no feature tracking was performed. It is hard to
visually follow regions and measure their energy change. Feature tracking
helps understand the energy transfer mechanisms in these vortex interaction
(energy cascade). It is also helpful for data reduction, data base management,
and model building.
In this animation, feature are extracted using segmentation based
upon thresholding, and they are followed over their lifetime. Objects are
colored based on the tracking result. The color of an object is inherit
from its parent (with the larger volume in the case of amalgamation).
In the first animation, one object is isolated and the rest are colored
in grey. In the second animation, one object is isolated and the rest are
removed. The change in mass and volume are shown below.
In this animation, objects are rendered with the volume rendering technique
to reveal the inner detail. The rainbow colormap indicates the variation
of the scalar value inside each object.
Statistics
Once objects have been tracked, we can also track properties of the
objects, such as mass, volume, moments, etc.
Image Gallery
Applications
Geophysics
Inlet Design
NCAR Weather Simulation
IBM Weather Forecast
EPA Weather Simulation
Plume
Isotropic
Turbulent Decay Simulation
Oceanography
The Shock-Elliptical Interaction - It has been adapted to 2D dataset!
Related Publications
- Tracking Scalar Features in Unstructured Datasets
- Deborah Silver and
Xin Wang
- Proceedings of IEEE Visualization 1998
- Tracking and Visualizing Turbulent
3D Features
- Deborah Silver and
Xin Wang
- IEEE Transaction on Visualization and Computer Graphics, Volume
3, No 2, June 1997
- Visualizing Evolving Scalar Phenomena
- Deborah Silver and
Xin Wang
- Invited Paper, Journal of Future Generations of Computer System, 586, 1998
- Volume Tracking
- Deborah Silver and
Xin Wang
- Proceedings of IEEE Visualization 1996, San Francisco, CA, 1996
- Octree-based Algorithm for 3D Feature
Tracking
- Xin Wang and Deborah
Silver
- CAIP Technical Report TR-204, CAIP Center, Rutgers University, 1995
Acknowledgment
The work is supported by DARPA HPCD
(DABT-63-93-C-0064), DOE (DE-FG02-93ER25179.A000),
NASA (NAG 2-829), and the CAIP
Center.
Xin Wang and
Prof. Deborah Silver
Department of Electrial and Computer Engineering
You are visitor # 
since Jan 28, 1998.
