Download presentation
Presentation is loading. Please wait.
1
January 5, 2006 1 Feature Tracking in VR for Cumulus Cloud Life-Cycle Studies E. J. Griffith, F. H. Post, M. Koutek, T. Heus and H. J. J. Jonker 11 th Eurographics Workshop on Virtual Environments Eric Griffith
2
January 5, 20062 Motivation – Data Visualization “The purpose of computation is insight, not numbers.” - Richard Hamming Insight is facilitated by a human in the loop Data visualization puts the human in the loop -0.259-0.5-0.707-0.866-0.966 -0.966-0.866-0.707-0.5-0.2590 0.2590.50.7070.8660.9661 0.8660.7070.50.2590
3
January 5, 20063 Motivation – Virtual Reality Why data visualization in VR? An extra visible dimension Improved perception High interactivity Dynamic viewpoint Dynamic data view Simplified, direct interaction with 3D data
4
January 5, 20064 Overview 1)Project Overview 2)Data Preprocessing 3)Interactive Visualization 4)Results 5)Conclusions
5
January 5, 20065 Overview I.Project Overview II.Data Preprocessing III.Interactive Visualization IV.Results V.Conclusions
6
January 5, 20066 Project Overview Cumulus cloud studies Largest source of uncertainty in climate models Use LES to explore cloud behavior in 3D Provide a clearer description of cloud dynamics Exploration in VR Interactive visualization of very large, time dependent data sets Meaningful visual representations of multivariate, 3D, time-dependent data sets
7
January 5, 20067 Large Eddy Simulation - LES Provides insight into flows in the Atmospheric Boundary Layer Solves for large-scale flows models small scale flows Used to simulate atmospheric conditions over tropical ocean Typically 6.4 x 6.4 x 3.2 km domains with 2 second time steps Grid sizes of 256 3 produce 112 GB of data for 1 simulated hour
8
January 5, 20068 Cloud Life-Cycles What is the “typical” cloud life-cycle? How do the clouds behave in each stage of the life-cycle? What causes the transition between stages?
9
January 5, 20069 The First Step – Cloud Selection Suitable clouds must be selected for study The challenges: Very large data sets Unpredictable and dynamic cloud behavior Qualitative selection criteria
10
January 5, 200610 Overview I.Project Overview II.Data Preprocessing III.Interactive Visualization IV.Results V.Conclusions
11
January 5, 200611 Data Preprocessing The objective of preprocessing is to convert the data into a format that can be meaningfully visualized interactively It is divided into two stages: Cloud tracking Isosurface creation
12
January 5, 200612 Cloud Tracking Cloud tracking involves the detection and tracking of features, which are clouds Tracking is simplified through two key observations: Clouds move and develop slowly Clouds that merge or split are either the same cloud or uninteresting to atmospheric scientists Detection is combined with tracking, and they are done via a 4D connected components algorithm
13
January 5, 200613 Cloud Tracking
14
January 5, 200614 Cloud Tracking
15
January 5, 200615 Cloud Tracking
16
January 5, 200616 Cloud Tracking
17
January 5, 200617 Cloud Tracking
18
January 5, 200618 Cloud Tracking
19
January 5, 200619 Isosurface Creation Isosurface creation transforms the raw data into visually meaningful representations of the clouds The creation follows a 5 stage pipeline process
20
January 5, 200620 Isosurface Pipeline 1.Prepare the data for isosurface creation 2.Create initial triangle meshes with Marching Cubes 3.Identify which meshes correspond with which clouds 4.Refine the meshes with a series of filters 5.Convert the meshes into triangle strips
21
January 5, 200621 Isosurface Pipeline 1.Prepare the data for isosurface creation 2.Create initial triangle meshes with Marching Cubes 3.Identify which meshes correspond with which clouds 4.Refine the meshes with a series of filters 5.Convert the meshes into triangle strips
22
January 5, 200622 Isosurface Pipeline 1.Prepare the data for isosurface creation 2.Create initial triangle meshes with Marching Cubes 3.Identify which meshes correspond with which clouds 4.Refine the meshes with a series of filters 5.Convert the meshes into triangle strips
23
January 5, 200623 Isosurface Pipeline 1.Prepare the data for isosurface creation 2.Create initial triangle meshes with Marching Cubes 3.Identify which meshes correspond with which clouds 4.Refine the meshes with a series of filters 5.Convert the meshes into triangle strips
24
January 5, 200624 Isosurface Pipeline 1.Prepare the data for isosurface creation 2.Create initial triangle meshes with Marching Cubes 3.Identify which meshes correspond with which clouds 4.Refine the meshes with a series of filters 5.Convert the meshes into triangle strips
25
January 5, 200625 Overview I.Project Overview II.Data Preprocessing III.Interactive Visualization IV.Results V.Conclusions
26
January 5, 200626 Cloud Explorer Cloud Explorer is our prototype of a VR visualization environment for cloud data It has been designed to enable atmospheric scientists to identify suitable clouds for study
27
January 5, 200627 Cloud Explorer Cloud Explorer components: a)Cloud field b)Volume graph c)World-in-Miniature d)Buttons e)Time control panel
28
January 5, 200628 Cloud Explorer
29
January 5, 200629 Overview I.Project Overview II.Data Preprocessing III.Interactive Visualization IV.Results V.Conclusions
30
January 5, 200630 Results – Data Preprocessing The preprocessing results are in terms of cpu time, output size and composition, and data compression The data sets were processed twice: once for all clouds and once for “complete” clouds 254.16 GB 42.36 GB (256 x 256 x 160) x 2169 Total Sizeq l SizeGrid Size x Time Steps 1011 MB1600 MB2h 4m 58s 1836 MB2557 MB11h 38m 22s Avgerage Memory Usage Peak Memory Usage CPU Time 64 : 1384 : 1679 MB46,824,994 6 : 136 : 16.9 GB517,110,152 Data Compression Ratios (vs q l ) Output Size Number of Triangles
31
January 5, 200631 Results – Cloud Explorer Depending on the number and size of visible clouds, Cloud Explorer provides frame rates between 20 and 30 FPS in stereo Currently, a session with Cloud Explorer results in the set of grid cells for each interesting cloud This data is then post processed to analyze the properties of the selected clouds “Fluffiness”, velocity profiles, mass flux, etc
32
January 5, 200632 Results – Cloud Mass Flux
33
January 5, 200633 Overview I.Project Overview II.Data Preprocessing III.Interactive Visualization IV.Results V.Conclusions
34
January 5, 200634 Conclusions Data preprocessing can provide sufficient data reduction to allow interactive VR visualization VR visualization enables scientists to combine theoretical and observational considerations Feature tracking with VR visualization can help scientists make sense out of very large, time- dependent data sets In depth studies of cloud behavior are made possible through cloud selection in VR
35
January 5, 200635 Future Directions The primary goal is to increase the amount and quality of time atmospheric scientists can spend using Cloud Explorer for cloud studies Large data handling facilities (e.g. out-of-core and multiresolution) for arbitrarily sized data sets Enhanced selection and interaction methods Visual information about multiple variables More context information about selected clouds
36
January 5, 200636 Acknowledgements Netherlands Organization for Scientific Research Netherlands National Computer Facility at SARA René Molenaar
37
January 5, 200637 Questions
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.