Presentation is loading. Please wait.

Presentation is loading. Please wait.

S. Mangiagli1, L. Drago2, S. Coco2, A. Laudani2, L. Lodato1, G

Similar presentations


Presentation on theme: "S. Mangiagli1, L. Drago2, S. Coco2, A. Laudani2, L. Lodato1, G"— Presentation transcript:

1 VTA: GRID based Thermal Images Processing for volcanic activity monitoring
S. Mangiagli1, L. Drago2, S. Coco2, A. Laudani2, L. Lodato1, G. Pollicino3 O.Torrisi1 1INGV, sezione di Catania 2DIEES, Università di Catania 3Consorzio COMETA Catania, February 2009

2 Volcanic Activity Monitoring by Thermal Images
A40 FLIR thermal camera installation between 25 and 27 June 2006 Map of Vulcano Island On June 2006 a A 40 thermal camera was installed in Vulcano La Fossa Crater. The location was in the internal and opposite crater flank (S1), 400 m distant from the fumarole field. The first two-year of data on temperature distribution frequency were recorded with this new methodology of acquisition, and automatically elaborated by software at INGV Catania Section. S1 Digital elevation model of ‘’La Fossa Crater’’

3 Volcanic Activity Monitoring by Thermal Images
A40 FLIR thermal camera installation between 25 and 27 June 2006 On June 2006 a A 40 thermal camera was installed in Vulcano La Fossa Crater. The location was in the internal and opposite crater flank (S1), 400 m distant from the fumarole field. The first two-year of data on temperature distribution frequency were recorded with this new methodology of acquisition, and automatically elaborated by software at INGV Catania Section. S1 Digital elevation model of ‘’La Fossa Crater’’

4 Volcanic Activity Monitoring by Thermal Images
A40 FLIR thermal camera installation between 25 and 27 June 2006 Thermal image On June 2006 a A 40 thermal camera was installed in Vulcano La Fossa Crater. The location was in the internal and opposite crater flank (S1), 400 m distant from the fumarole field. The first two-year of data on temperature distribution frequency were recorded with this new methodology of acquisition, and automatically elaborated by software at INGV Catania Section. Visible image

5 Methodology of data acquisition
Ambient temperature, relative humidity, DCF time, rainfall rate are acquired by the meteo station and directly put within the thermal camera software for applying the atmospheric corrections each 30 min. Thermal images and meteo data are acquired in our site in Lipari, their transfer and storage in Catania occurs thanks to a GARR connection. The thermal camera started to work in September 2006. The acquisition frequency is of a frame each 30 min. Workstation for the data saving and subsequent analysis

6 The Volcano Thermo Analysis (VTA) software
Data loading. ROI definition. Temperature ranges definition. Spatial, temporal and statistic analysis.

7 What kind of input data?

8 Data loading Thermal images listed by date and time
Selection of images belonging to the same: Day Week Month Year

9 Temperature ranges definition
Up to 10 different temperature ranges in order to filter data. Only points inside these ranges will be considered. Regions Of Interest (ROI) Unlimited number of ROIs. Rectangular, elliptical, punctiform, irregular roi. Possibility to save ROIs.

10 ROI visualization

11 ? What kind of output? Three kinds of output:
Spatial localization info. Temporal trend. Statistical info.

12 Spatial localization Single image: ROI indication. Ranges indication.
MAX and min values. Vertices height proportional to pixels temperature. Colorbar indicating temperature. Reference greyscale image.

13 Spatial localization Multiple images:
Average, MAX e min temperature graph. ROI indication. Ranges indication. Vertices height proportional to pixels temperature. Colorbar indicating temperature. Reference greyscale image.

14 Some examples

15 Some examples

16 Temperature trend Max and average temperature 2D plot, with one point for each: Image Hour Day Week Month Year

17 Frequency histograms Histograms showing temperature frequency (how many time is that value present?), with one histogram for each: Image Hour Day Week Month Year

18 Great! But just a moment…
X 1≈ 300 KB (76800 pixels). Let’s consider one image every 30 minutes… 1 day: 14 MB 1 week: 100 MB 1 month: 400 MB 1 year: 4,8 GB Is your machine powerful enough? YES? OK…

19 We need more images with better resolution…
…unfortunately… One image every 5 minutes: 1 month: 2.5 GB 1 year: 30 GB pixels with 640*480 image resolution. You want to perform analysis for 5 temperature ranges, 24 ROIs and 2 years of data. We need more images with better resolution… It’s up to you! Is still your machine powerful enough?

20 Porting VTA to the GRID VTA is developed using IDL (Interactive Data Language): IDL must be installed on the GRID. IDL must be used by command line. VTA has its own interface: It must be adapted to run in batch mode. VTA works with a huge number of images: Thermal images must be uploaded to the grid.

21 VTA as a normal JOB Ok...it works. But is it the best we can do?
JDL file: Type = "Job"; JobType = "Normal"; Executable = "start_idl.sh"; StdOutput = "std.out"; StdError = "std.err"; Arguments = "lfn:/grid/cometa/INGV/VTA/vtagrid.sav"; InputSandbox = {"start_idl.sh","settings.txt","roi.sav","ranges.sav"}; OuTputSandbox = {"output.tar.gz","std.err","std.out"}; Requirements = Member("IDL-6.4",other.GlueHostApplicationSoftwareRunTimeEnvironment); RetryCount = 5; Ok...it works. But is it the best we can do?

22 Do you know DAG? Graph Direct Acyclic Graph Direct Graph

23 VTA as a DAG

24 VTA as a DAG

25 VTA as a DAG [ type = "dag"; max_nodes_running=10;
VirtualOrganisation = "cometa"; InputSandbox = {"roi.sav","ranges.sav","settings.txt"}; nodes = [node1 = [description = [Executable = "start_idl1.sh"; Arguments = "lfn:/grid/cometa/INGV/VTA/vtagriddag.sav 1"; StdOutput = "std1.out"; StdError = "std1.err"; InputSandbox = {"start_idl1.sh",root.InputSandbox}; OutputSandbox = {"output1.tar.gz","std1.err","std1.out"}; Requirements = Member("IDL-6.4",other.GlueHostApplicationSoftwareRunTimeEnvironment); ]; node2 = [description = [Executable = "start_idl2.sh"; Arguments = "lfn:/grid/cometa/INGV/VTA/vtagriddag.sav 2"; StdOutput = "std2.out"; StdError = "std2.err"; InputSandbox = {"start_idl2.sh",root.InputSandbox }; OutputSandbox = {"output2.tar.gz","std2.err","std2.out"}; node3 = [description = [Executable = "start_idl3.sh"; Arguments = "lfn:/grid/cometa/INGV/VTA/aggregator.sav 2"; inputSandbox = { "start_idl3.sh", root.InputSandbox, root.nodes.node1.description.OutputSandbox[0], root.nodes.node2.description.OutputSandbox[0] }; StdOutput = "std.out"; StdError = "std.err"; OutputSandbox = {"output.tar.gz","std.err","std.out"}; dependencies = { {node1,node3},{node2,node3} }; ]

26 Job’s execution time is not perfectly predictable
VTA on GRID results 1 month of data on a PC = 20 minutes (Intel Core2Duo GHz, 2 GB RAM) 2 month of data on Grid (DAG - 2 nodes) = 15 minutes (average) 4 month of data on Grid (DAG - 2 nodes) = 40 minutes (average) Job’s execution time is not perfectly predictable since Total job’s execution time = Job’s handling time + effective running time.

27 Conclusions We adopted a new method for Volcanic Activity Monitoring by mean of Thermal Images, easily analized due to a specific software (VTA) developed in IDL. GRID greatly enhanced VTA’s potentialities, allowing us to perform faster and multiple analysis on huge set of data, proving itself as a really usefull instrument for scientific research.


Download ppt "S. Mangiagli1, L. Drago2, S. Coco2, A. Laudani2, L. Lodato1, G"

Similar presentations


Ads by Google