0
ON THE EXTENSION OF THE PRODUCT MODEL IN POLSAR PROCESSING FOR UNSUPERVISED CLASSIFICATION USING INFORMATION GEOMETRY OF C...
K-MEANS CLASSIFIER <ul><li>Conventional clustering algorithm: </li></ul><ul><li>Initialisation: Assign pixels to classes. ...
OUTLINE <ul><li>Non-Gaussian clutter model: the SIRV model </li></ul><ul><li>Contribution of the geometry of information <...
OUTLINE
CONVENTIONAL COVARIANCE MATRIX ESTIMATE <ul><li>With low resolution, clutter is modeled as a Gaussian process. </li></ul><...
SCM IN HIGH RESOLUTION Gamma classification Wishart classification with SCM Results are very close from each other : influ...
THE SIRV MODEL Non-Gaussian SIRV (Spherically Invariant Random Vector) representation of the scattering vector  : <ul><li>...
COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL ML ESTIMATE UNDER SIRV ASSUMPTION ...
DISTANCE BETWEEN COVARIANCE MATRICES UNDER SIRV ASSUMPTION <ul><li>Non Gaussian Process  ↔  Generalized LRT  ↔  SIRV dista...
COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL RESULTS ON REAL DATA Color composi...
OUTLINE
Euclidian Mean: CONVENTIONAL MEAN OF COVARIANCE MATRICES The mean in the Euclidean sense of  n  given positive-definite He...
Riemannian Mean: A DIFFERENTIAL GEOMETRIC APPROACH TO THE GEOMETRIC MEAN OF HERMITIAN DEFINITE POSITIVE MATRICES The mean ...
OUTLINE
CLASSIFICATION RESULTS Wishart classification with SCM, Arithmetical mean SIRV classification with FPE, Arithmetical mean ...
CLASSES IN THE H- α  PLANE
PARACOU, FRENCH GUIANA <ul><li>Acquired with the ONERA SETHI system </li></ul><ul><li>UHF band </li></ul><ul><li>1.25m res...
CLASSIFICATION RESULTS Classification with Wishart distance, Arithmetical mean Classification with Wishart distance, Geome...
OUTLINE
CONCLUSIONS <ul><ul><li>Further investigation of the distance is required. </li></ul></ul><ul><ul><li>Interpretation is di...
Upcoming SlideShare
Loading in...5
×

Formont.ppt

223

Published on

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

  • Be the first to like this

No Downloads
Views
Total Views
223
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
4
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • Gamma classification : no information relative to the polarimetric phase information: ONLY SPAN
  • Rajouter slide pour M chapeau. Eventuellement rajouter un slide pour présenter diff SCM et FP.
  • Citer papier de Lee
  • Interprétation difficile, premiers résultats
  • Transcript of "Formont.ppt"

    1. 1. ON THE EXTENSION OF THE PRODUCT MODEL IN POLSAR PROCESSING FOR UNSUPERVISED CLASSIFICATION USING INFORMATION GEOMETRY OF COVARIANCE MATRICES P. Formont 1,2 , J.-P. Ovarlez 1,2 , F. Pascal 2 , G. Vasile 3 , L. Ferro-Famil 4 1 ONERA, 2 SONDRA, 3 GIPSA-lab, 4 IETR
    2. 2. K-MEANS CLASSIFIER <ul><li>Conventional clustering algorithm: </li></ul><ul><li>Initialisation: Assign pixels to classes. </li></ul><ul><li>Centers computation: Compute the centers of each class as follows: </li></ul><ul><li>Reassignment: Reassign each pixel to the class whose center minimizes a certain distance. </li></ul>
    3. 3. OUTLINE <ul><li>Non-Gaussian clutter model: the SIRV model </li></ul><ul><li>Contribution of the geometry of information </li></ul><ul><li>Results on real data </li></ul><ul><li>Conclusions and perspectives </li></ul>
    4. 4. OUTLINE
    5. 5. CONVENTIONAL COVARIANCE MATRIX ESTIMATE <ul><li>With low resolution, clutter is modeled as a Gaussian process. </li></ul><ul><li>Estimation of the covariance matrix of a pixel, characterized by its target vector k , thanks to N secondary data: k 1 , …, k N . </li></ul><ul><li>Maximum Likelihood estimate of the covariance matrix, the Sample Covariance Matrix (SCM): </li></ul>
    6. 6. SCM IN HIGH RESOLUTION Gamma classification Wishart classification with SCM Results are very close from each other : influence of polarimetric information ?
    7. 7. THE SIRV MODEL Non-Gaussian SIRV (Spherically Invariant Random Vector) representation of the scattering vector : <ul><li>where is a random positive variable (texture) and (speckle). </li></ul><ul><ul><li>The texture pdf is not specified : large class of stochastic processes can be described. </li></ul></ul><ul><ul><li>Texture : local spatial variation of power. </li></ul></ul><ul><ul><li>Speckle : polarimetric information. </li></ul></ul><ul><ul><li>Validated on real data measurement campaigns. </li></ul></ul>
    8. 8. COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL ML ESTIMATE UNDER SIRV ASSUMPTION <ul><ul><li>Under SIRV assumption, the SCM is not a good estimator of M . </li></ul></ul><ul><ul><li>ML estimate of the covariance matrix: </li></ul></ul><ul><ul><li>Existence and unicity. </li></ul></ul><ul><ul><li>Convergence whatever the initialisation. </li></ul></ul><ul><ul><li>Unbiased, consistent and asymptotically Wishart-distributed. </li></ul></ul>
    9. 9. DISTANCE BETWEEN COVARIANCE MATRICES UNDER SIRV ASSUMPTION <ul><li>Non Gaussian Process ↔ Generalized LRT ↔ SIRV distance between the two FP covariance matrices </li></ul><ul><li>Gaussian Process ↔ Generalized LRT ↔ Wishart distance between the two SCM covariance matrices </li></ul>
    10. 10. COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL COVARIANCE MATRIX ESTIMATE : THE SIRV MODEL RESULTS ON REAL DATA Color composition of the region of Brétigny, France Wishart classification with SCM Wishart classification with FPE
    11. 11. OUTLINE
    12. 12. Euclidian Mean: CONVENTIONAL MEAN OF COVARIANCE MATRICES The mean in the Euclidean sense of n given positive-definite Hermitian matrices M 1 ,.., M n in P ( m ) is defined as: Barycenter:
    13. 13. Riemannian Mean: A DIFFERENTIAL GEOMETRIC APPROACH TO THE GEOMETRIC MEAN OF HERMITIAN DEFINITE POSITIVE MATRICES The mean in the Riemannian sense of n given positive-definite Hermitian matrices M 1 ,.., M n in P ( m ) is defined as: Geodesic: Riemannian distance:
    14. 14. OUTLINE
    15. 15. CLASSIFICATION RESULTS Wishart classification with SCM, Arithmetical mean SIRV classification with FPE, Arithmetical mean SIRV classification with FPE, Geometrical mean
    16. 16. CLASSES IN THE H- α PLANE
    17. 17. PARACOU, FRENCH GUIANA <ul><li>Acquired with the ONERA SETHI system </li></ul><ul><li>UHF band </li></ul><ul><li>1.25m resolution </li></ul>
    18. 18. CLASSIFICATION RESULTS Classification with Wishart distance, Arithmetical mean Classification with Wishart distance, Geometrical mean Classification with geometric distance, Geometrical mean
    19. 19. OUTLINE
    20. 20. CONCLUSIONS <ul><ul><li>Further investigation of the distance is required. </li></ul></ul><ul><ul><li>Interpretation is difficult because no literature. </li></ul></ul><ul><ul><li>Span can give information for homogeneous areas. </li></ul></ul><ul><ul><li>Necessity of a non-Gaussian model for HR SAR images. </li></ul></ul><ul><ul><li>Geometric definition of the class centers in line with the structure of the covariance matrices space. </li></ul></ul>
    1. A particular slide catching your eye?

      Clipping is a handy way to collect important slides you want to go back to later.

    ×