Group Tracking in Scientific Visualization
Transcript of Group Tracking in Scientific Visualization
Group Tracking in Scientific Visualization Sedat OZER1, Deborah SILVER1, Pino MARTIN3
1CAIP, Visualization Lab, Dept. of Electrical & Computer Engineering, Rutgers University, NJ 3CROCCO Lab, Dept. of Aerospace Engineering, Rutgers, University of Maryland, MD
Sedat Ozer: [email protected]
Deborah Silver: [email protected]
Pino Martin: [email protected] Rutgers University
We gratefully acknowledge the support of SciDAC Institute for Ultra-Scale Visualization, http://vis.cs.ucdavis.edu/Ultravis/, DOE #DE-FG02-09ER25977
Petri Net
1. Overview 5. Group Tracking Model
Packet _B:
(Side View)
Tracking Packet_A (Side View)
t0 t1 t2 t3 t4 t5
z
z
x α
Where α<450
Group_S Group_R
Group_R Group_S
. . .
...
Feature Segmentation Attribute Computation Feature tracking Feature based visualization
1st level (feature) extraction module
Feature Segmentation Attribute Computation
Group tracking (choose the groupID that
gives the max feature volume among the
candidate groupIDs)
Feature based visualization
2nd level (group) extraction module
ti ti + ti-1
ti + ti-1 ti
Feature Segmentation Attribute Computation Higher level group tracking Feature based visualization
nth level (higher level) group extraction module
ti + ti-1 ti Time step j+1
Time step j
(I) Feature Tracking at t0. In feature tracking each feature has a unique color. (A total of 262 features at t0)
(II) Group Tracking at t0. In group tracking each group has a unique color. (A total of 177 packets at t0)
Feature_a Feature_b Feature_a
Feature_b
As simulations increase in size and complexity, the number of
coherent features found in the data also increases. While feature
tracking follows the evolution of individual structures, the interaction
and movements of groups of features has not been fully addressed.
Identifying and visualizing such groups and modelling their
complicated interactions are important in many domains. In this work,
we propose a group tracking model to track and follow groups of
features interacting together. We demonstrate our model on a 3D
time-varying simulation of a wall bounded turbulent flow.
Tracking Feature_a (Side View)
t0
t1
t2
t3
t4
t5
7. More Results
4. Group Events: An illustration
6. Application in Computational Fluid Dynamics (CFD): Wall Bounded Turbulence Direct Numerical Simulation (DNS) Data
• Packet (Illustrated above): Each feature (yellow) of the group
elongates around a fluid drop (blue) to make a certain shape. The
angle (α) between the max values of the two closest features
should be smaller than 450, while they remain close enough to
each other. (i.e. within a predefined distance along x, y and z).
• Super-structure (Illustrated below): A group of packets that move
coherently.
t0
t1
t2
t3
t4
t5
t0 t1 t2 t3 t4
t5
t0
t1 t2 t3 t4 t5
x
• Features moving coherently form groups.
• Group tracking provides a new set of
information related to the group-feature
relations.
• Group examples:
• A set of stars or halos in cosmology,
• A set of hairpin vortices in CFD,
• A set of clouds in a storm formation.
3. Group Examples
Group
Feature
Galaxy
Storm
Packet
Vortex
Cloud
Star
2. Challenges
• Definition of a group changes from domain to domain. This variety
of definitions makes it harder to find a common and efficient
(automatic) technique to group features in different domains.
• Defining merge or split events is complicated at group level than at
the feature level.
• Multi-level structure brings new events to be defined in each
domain, such as cross (group-feature) level events in CFD
simulations (e.g. where a feature changes its group).
• Tracking groups (the correspondence problem for the groups).
• Groups have the same set of events that features do. Besides,
they also have group-feature (cross) level events.
• The figure shows that one feature from Group_R moves into
Group_S (a cross level event) without performing any primitive
events from time step j to j+1. This event can be detected by
using the group tracking model.
Packet_A
Packet_B
An illustration of a super-structure in wall bounded turbulence