Your SlideShare is downloading. ×
0
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Ieee gold angiati
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Ieee gold angiati

387

Published on

Published in: Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
387
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
7
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. IEEE Gold Remote Sensing Conference 2010
    Naval Academy, Livorno, Italy, April 29-30, 2010
    Flooding Maps From Cosmo-Skymed Images
    Elena Angiati
    Silvana Dellepiane
    University of Genoa (Italy)
    Dept. of Biophysical and Electronic Engineering (DIBE)
    NUMIP – NUMerical Image Processing
  • 2. Outline
    Introduction:
    Identification of flooded areas;
    The proposed method:
    fast-ready flood maps  pre-processing & RGB composition;
    detailed flood maps segmentation approach.
    Experimental results:
    experiments on SAR images
    Conclusions
    “OPERA – Civil protection from floods” pilot project - Italian Space Agency & Italian Department for Civil Protection.
  • 3. Introduction
    Multitemporal remote-sensing images represent a powerful source of information for monitoring the evolution of the Earth’s surface
    Relevant task: identification of flooded areas.
    SAR images are particularly useful during floods:
    all-weather capability
    cloud-penetrating properties
  • 4. Fast-ready flood maps
    An RGB composition is used, where two images are combined into a false colour composite image  enhancing the flooded areas
    Images can be acquired with different sensor parameters  an appropriate pre-processing is required
    Three sequential steps are proposed:
    filtering,
    adaptive histogram truncation,
    equalization.
  • 5. Filtering
    Comparison of different filters: SRAD (Speckle Reducing Anisotropic Diffusion), Lee, Frost, Enhanced Lee and Frost filters.
    SRAD allows to reduce noise and to preserve details.
    Best performances in the frequency domain  mean preservation and isotropic behavior.
    Original image Lee Frost
    Enhanced Lee Enhanced Frost SRAD
  • 6. Histogram Equalization
    Linear shrinking from 2 Bytes to 1 Byte  loss of many informative contents, due to the very long distribution tail
    Histogram equalization  normalization of the different histogram distributions
    Usual histogram equalization is not properly working with such a heavy tail.
     Adaptive histogram truncation is applied
    Zoom into the interval 0-500 of original histogram of image (maximum value = 18000)
    Histogram of equalized image
  • 7. Histogram truncation & Equalization
    Preliminary clipping to the 95th percentile & equalization  best performances
    RGB composition image is obtained:
    Red channel: difference between pre and post-event
    Green: post-event image
    Blue: pre-event image
    Blue = uniformcumulative function Magenta = cumulative ofimage
    Adaptivehistogramequalization (truncation& equalization)
    Histogramequalizationoforiginalimage
  • 8. Detailed flood maps
    A multi-seed-growing segmentation approach is employed.
    Segmentation process:
    uses filtered images;
    starts from water pixels;
    uses an anisotropic image-scanning mechanism  order of pixel analysis is dependent on the image content.
    Test rule  a similarity criterion is satisfied.
  • 9. Segmentation algorithm
    Given the seed point , a “seed region” is generated, using the seed pointand its direct 8-neighbours:
    The sample mean is computed:
    Sample standard deviation is computed on a 5x5 window
    centered on the seed pixel:
  • 10. Segmentation algorithm
    Sample mean m aggregation rule
    Sample standard deviation s estimate the threshold value.
    The threshold is adaptive to the scattering of the region of interest and is set to:
    A new pixel is assigned to the region if its distance with respect to the “seed region” is small enough.
  • 11. Data set
    Different multitemporal data set consisting of pair of co-registered Cosmo/Skymed images are used.
    Flood event of the MassaciuccoliLake: images in Stripmap acquisition modes, with different geometric acquisition parameters
    Cosmo/SkymedStripmap images (spacial resolution: 2,5 meters)
    LEFT: 20th December 2009 (ascending/right looking angle)
    RIGHT: 30th December 2009 (descending/left looking angle)
  • 12. Example of Fast-ready flood map
    The images could be used in an RGB composition despite the different acquisition parameters
    RGB composition.
    In magenta: change due to decrease of backscattering, corresponding to flooded areas.
    In cyan: no-change due to high backscattering in both images
    In bordeaux: no-change due to low backscattering in both images
  • 13. Example of Detailed flood map
    The segmentation process is not affected by different acquisition setting  the filtered images can be used.
    Detailed map of flooded areas.
    In blue: steadywater
    In cyan: flooded areas
  • 14. Otherexamples on Stripmapimages
    Cosmo/Skymed images acquired near Scutari (Albania) in Stripmap mode (spatial resolution: 2,5 meters), with different acquisition parameters
    10th January 2010 - in descending configuration with right look angle
    15th January 2010 - in ascending configuration with right look angle
    Fast-readyfloodmap
    Detailedfloodmap
    Floodedareas
    Floodedareas
    Steady water
    Steady water
    No floodedareas
    Otherchanges
  • 15. Otherexamples on Stripmapimages
    Cosmo/Skymed images acquired near Alessandria (Italy) in Stripmap mode (pixel resolution: 2,5 meters), in descending configuration withright look angle
    30th April 2009
    1st May 2009
    Detailedfloodmap
    Fast-readyfloodmap
    Floodedareas
    Floodedareas
    No floodedareas
    Steady water
    Steady water
    Otherchanges
  • 16. Examples on Spotlightimages
    Cosmo/Skymed images acquired near Alessandria (Italy) in Spotlight mode (pixel resolution: 0,5 meters), with different acquisition parameters
    1st May 2009 - in ascending configuration with right look angle
    29th April 2009 - in descending configuration with left look angle
    30th April 2009 - in ascending configuration with right look angle
    Fast-readyfloodmap
    Detailedfloodmap
    Multitemporalfloodmap
    Floodedareas at 29th April 2009
    Floodedareas at 30th April 2009
    Steady water
    Floodedareas
    No floodedareas
    Floodedareas
    Steady water
    Steady water
    Otherchanges
  • 17. Conclusions
    Several image processing techniques and a segmentation method have been proposed.
    Images acquired by the new mission Cosmo/Skymed have been used for experiments.
    Both qualitative and quantitative algorithms have been presented and very good performances have been obtained in both cases.

×