Coherent Change Detection Esfahan, Iran - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Coherent Change Detection Esfahan, Iran

Description:

Take mean shift of remaining patches and apply to entire image ... Quickbird panchromatic. 6. Compute coherence between pairs of images ... – PowerPoint PPT presentation

Number of Views:224
Avg rating:3.0/5.0
Slides: 12
Provided by: regi114
Category:

less

Transcript and Presenter's Notes

Title: Coherent Change Detection Esfahan, Iran


1
(No Transcript)
2
Coherent Change Detection Esfahan, Iran
  • Alex Rodrigues
  • GMOSS WP20300 meeting at Royal Military Academy,
    Brussels
  • 14 November 2005

3
Data
  • RADARSAT fine beam coherent triplet
  • 20 September 2002
  • 05 July 2003
  • 29 July 2003
  • Quickbird optical data
  • 15 June 2004

4
Co-registration
  • Coherent SAR
  • frozen speckle pattern
  • coregistered SAR images using localised cross
    correlation of speckle patterns.
  • compute cross-correlation over 49 small patches.
  • Reject incoherent patches.
  • Take mean shift of remaining patches and apply to
    entire image
  • Coregistration accurate to 0.15 pixels
  • Coregistration of 3 images was done as follows
  • Earliest is master
  • Middle image is coregistered to image 1
  • Latest image is coregistered to image 1

5
Coregistered SAR data Esfahan site
RADARSAT Colour Composite
Quickbird panchromatic
20 September 2002 5 July 2003 29 July 2003
6
Coherence
  • Compute coherence between pairs of images
  • Similarity between pairs of images
  • Several possibilities
  • Earliest and middle - long time gap (10 months)
  • Earliest and latest (not done here)
  • Middle and latest - short time gap ( 1 month)
  • Produce visual products
  • Colour composites of SAR images and coherence
  • Earliest, middle, coherence
  • Middle, latest, coherence
  • Visually inspect products

7
Colour composites of SAR images and coherence
  • Earlier SAR image 1 red
  • Later SAR image 2 green
  • Coherence between image 1 and image 2 - blue
  • The colour scheme is as follows
  • Yellow change to existing objects (eg building
    modification, object rearrangement)
  • White and blue no change
  • Red demolition, road building
  • Green - construction

8
20/09/2002, 05/07/2003, coherence
Middle
Earliest
9
05/07/2003,29/07/2003, coherence
Middle
Latest
Much shorter time interval natural areas have
high coherence
10
Conclusions
  • Changes have been detected
  • hard to spot requires a trained eye (and some
    guesswork!)
  • RADARSAT resolution is not high enough
  • Reliable change detection requires continuous
    monitoring with regular repeats (so that changes
    between images of fixed temporal difference are
    compared)
  • Coherent image acquisition places stringent
    requirements on satellite orbit control
  • Would like to try to create some rules and try to
    classify the SAR/coherence composites
  • Fusion with optical data

11
(No Transcript)
Write a Comment
User Comments (0)
About PowerShow.com