Your SlideShare is downloading. ×
Recent Advances in Object-based Change Detection.pdf
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

Recent Advances in Object-based Change Detection.pdf


Published on

Published in: Technology, Business

  • Be the first to comment

  • Be the first to like this

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

No notes for slide


  • 1. Mitglied der Helmholtz-Gemeinschaft IGARSS 2011, Vancouver Change Detection and Multitemporal Image Analysis I Recent Advances in Object-based Change Detection July 25, 2011 | Irmgard Niemeyer, Clemens Listner Nuclear Safeguards Group Institute of Energy and Climate Research IEK-6: Nuclear Waste Management and Reactor Safety Forschungszentrum Jülich GmbH, Germany
  • 2. AcknowledgmentsGerman Support Programme for theInternational Atomic Energy Agency (IAEA)Project on satellite imagery analysis and photointerpretation support“EC FP7, Global Monitoring for Environment andSecurity (GMES)Current project G-MOSAICGeneral R&D interestsMethodological developments, PhD thesis Listner Slide 2
  • 3. Recent Advances in Object-basedChange Detection Slide 3
  • 4. Very high spatial resolution opticalsensors (<1m): WorldView-2 Slide 4
  • 5. Object-based change detection usingIR-MAD Iteratively Reweighted Multivariate Alteration Detection (IR-MAD) [Nielsen 2007] Linear transformation of the feature space aimed to enhance the change information in the difference image Modeling object’s feature vector as random vectors F and G of length N Transformation of vectors to enhance relevant changes var(m1 = a1TU - b1TV) → max under the constraint that var(a1TU) = var(b1TV) = 1 Further orthogonal variates mi can be computed Σmi2 ~ Chi2 indicating change probability P(change) Iteration by weighting with 1- P(change) Additional step: Application of PCA to U and V1. Introduction Slide 5
  • 6. Object-based change detection usingIR-MAD Statistical pixel-based change detection approaches provide good results, but shows limits due to … • low number of spectral channels or small spectral range covered, • image registration problems. Object-based change detection looks promising, but … • how to connect corresponding objects? • how to carry out a reasonable segmentation for this task?1. Introduction Slide 6
  • 7. Existing approaches to segmentationfor object-based change detection Segment I1 and I2 as stack Time 1 • segmentation not adequate for I1 Segmentation and I2 Time 2 levels Image data • shape features cannot be used Use segmentation of I1 for I2 Time 1 • segmentation not adequate for I2 Time 2 Segmentation • shape features cannot be used Image data levels Independent segmentation Time 1 • leads to false-alarm segment changes Time 2 Segmentation levels • shape features can be used Image data2. Segmentation Slide 7
  • 8. Multiresolution segmentation Region-based bottom-up approach to segmentation Each segment is a binary tree (leafs=pixel, root=final segment) Implemented in eCognitionTM Starts with chessboard segmentation Selects iteratively a segment X and merges it to a neighboring segment Y if X  (( , d , ) min)) ( Y d Z X  Z () NX Y  ((, d, ) min)) (X dZ Y  Z () NY d( , X)T Y2. Segmentation Slide 8
  • 9. Multiresolution segmentation2. Segmentation Slide 9
  • 10. Multiresolution segmentation applied toslightly different imagesSegmentation of identical images up to Gaussian noise (μ=0,σ=0.1) usingmultiresolution segmentation2. Segmentation Slide 10
  • 11. Multiresolution segmentation adaptedfor object-based change detection 11. Segment I1 using multiresolution segmentation2. Apply this segmentation to I2 and recalculate color heterogeneity3. Check each merge for consistency with I2 using a predefined test4. Remove inconsistent segments using a predefined removal strategy5. Re-run the multiresolution segmentation using the so gained segmentation of I2 as an initial segmentation2. Segmentation Slide 11
  • 12. Multiresolution segmentation adaptedfor object-based change detection 2 Given segment S3 with children S1 (seed) and S2 Threshold test • h(S3) ≤ Tcheck in I2 ? Local best fitting test • Is S2 the locally best fitting neighbor for S1 in I2 ? Local mutual best fitting test • Are S1 and S2 local mutually best fitting in I2 ? Reduce sensitivity of the best fitting tests by using Tchecktolerance2. Segmentation Slide 12
  • 13. Segmentation for object-basedchange detectionThreshold test & universal segment removal strategy2. Segmentation Slide 13
  • 14. Segmentation for object-basedchange detectionLocal mutual best fitting test & global segment removalstrategy2. Segmentation Slide 14
  • 15. Segmentation for object-basedchange detectionLocal best fitting test & local segment removal strategy2. Segmentation Slide 15
  • 16. Segmentation for object-basedchange detectionThreshold test & universal segment removal strategy2. Segmentation Slide 16
  • 17. Object correspondence for object-based change detection Directed Via intersection xi = f x  Si  , xi = f x  S1  , 1 n yi = f y  S 2  yi =  f y Tk  n k=13. Object correspondence Slide 17
  • 18. Object-based change detection Pre-processing Image-to-image registration, Radiometric normalization Canty & Nielsen 2009 Segmentation Multiresolution segmentation adapted e.g. Listner & Niemeyer 2010, 2011a, to change detection 2011b Change detection Nielsen 2007, Listner & Niemeyer IR-MAD 2011b Change classification Class-based FFN Marpu 2009 Post-processing Integration to GIS or GDBS4. Experiments Slide 18
  • 19. Object-based change detection4. Experiments Slide 19
  • 20. Object-based change detectionSegmentation of the bitemporal imagery using threshold test anduniversal segment removal strategy.4. Experiments Slide 20
  • 21. Object-based change detectionDirected change detection. Changes from time 1 to time 2 (left) andfrom time 2 to time 1 (right).4. Experiments Slide 21
  • 22. Object-based change detectionChange detection using intersected Change detection using MAD objects. objects.4. Experiments Slide 22
  • 23. Object-based change detectionAccuracy assessment Directed change Directed change Change Change detection: detection: detection using detection using T1T2 T2T1 intersected MAD objects objectsOverall 0.98 0.98 0.98 0.99accuracyKIA 0.82 0.87 0.77 0.754. Experiments Slide 23
  • 24. Summary An enhanced procedure for segmentation was introduced and implemented into the change detection workflow. Moreover, numerically issues in the IR-MAD method were addressed. The proposed methods showed good results in three experiments using aerial imagery. Further developments are needed: • New consistency tests and segment removal strategies; • methods for enabling the user to easily select the segmentation parameters, e.g. by using training samples; • implementation as eCognition plugin.5. Summary Slide 24
  • 25. Most recent publications C. Listner and I. Niemeyer (2011a), “Advances in object- based change detection,” Proc. IGARSS 2011, Vancouver, July 2011 C. Listner and I. Niemeyer (2011b), “Object-based change detection,” Photogrammetrie, Fernerkundung, Geoinformation (PFG), vol. 3, 2011 (in print) Slide 25
  • 26. Thank you for your attention. Dr. Irmgard Niemeyer Nuclear Safeguards Institute of Energy and Climate Research IEK-6: Nuclear Waste Management and Reactor Safety Forschungszentrum Jülich GmbH in der Helmholtz-Gemeinschaft | 52425 Jülich | Germany Phone / Fax: +49 2461 61-1762 / -2450 Email: Slide 26