Multiscale Detection and Characterisation of CMEs

1
Multiscale Detection and Characterisation of CMEs

• J. P. Byrne1, P. T. Gallagher1, C. A. Young2 and
• R. T. J. McAteer3
• 1 Astrophysics Research Group, School of Physics,
  Trinity College Dublin, Dublin 2, Ireland.
• 2 ADNET Systems Inc., NASA Goddard Space Flight
  Center, Greenbelt, MD 20771, USA.
• 3 Catholic University of America, NASA Goddard
  Space Flight Center, Greenbelt, MD 20771, USA.

STEREO/Cor1 24-Jan-07
2
Overview

• CME Models
• Image Processing Multiscale methods
• CME Morphology Kinematics
• Application to LASCO and STEREO/SECCHI

LASCO/C3 27-Feb-00
3
CME Models

• Magnetic Flux-Rope

• Forbes Priest, 1990
• Chen Krall, 2003

• Magnetic Break-out

• Antiochos et al. 1999
• Lynch et al. 2004

4
Image Pre-Processing

• Normalisation
• - exposure time
• - CCD bias
• - data dropouts
• Background subtraction
• Median filtering
• (de-noising)

LASCO/C2 18-Apr-00
5
Finding the CME Front

• Edge Detection

6
Our Algorithm
Image Pre-Processing
1) Multiscale Decomposition
Vector-Arrow Field
2) Gradient Space Information
3) Spatio-Temporal Filter
4) Non-Maxima Suppression
5) CME Front Characterisation
Kinematics Morphology
7
Our Algorithm
Image Pre-Processing
1) Multiscale Decomposition
CME Vector Flow Field
2) Spatio-Temporal Filter
3) Non-Maxima Suppression
4) CME Front Characterisation
Kinematics Morphology
8
1) Multiscale Decomposition

• Wavelets
• Scaling / dilation factor (a)
• Shifting / translation factor (b)
• Suppress noise

9
1) Multiscale Decomposition
Input f
Low pass Approximation
High pass Detail
LASCO/C2 24-Jan-07
10
1) Multiscale Decomposition
Horizontal Direction
Scale 1
Scale 3
Scale 5
Vertical Direction
Scale 1
Scale 3
Scale 5
11
2) Gradient Space Information
Consider the Gradient of an image(which points
in the direction of most rapid change)
The gradient specifies
1) Magnitude
2) Direction
We have the Detail at a scale (N1) resulting
from the directional Derivative-of-Gaussian
convolved with the Approximation at scale (N)
So too can the Magnitude and Direction be taken
from the multiscale decomposition as illustrated
12
2) Gradient Space Information
Original
Magnitude
Angle
(McAteer et al. 2007)
13
2) Gradient Space Information
Vectors with magnitude and inclination angle
14
Our Algorithm
Image Pre-Processing
1) Multiscale Decomposition
CME Vector Flow Field
2) Spatio-Temporal Filter
3) Non-Maxima Suppression
4) CME Front Characterisation
Kinematics Morphology
15
2) Spatio-Temporal Analysis
Degrees of Freedom Scale, Magnitude Angle
in Space Time
16
Our Algorithm
Image Pre-Processing
1) Multiscale Decomposition
CME Vector Flow Field
2) Spatio-Temporal Filter
3) Non-Maxima Suppression
4) CME Front Characterisation
Kinematics Morphology
17
3) Non-Maxima Suppression

• Nearest-neighbour info.
• Criteria of angle and magnitude from gradients.
• Pixels chained along edges.

(Image by C.A.Young)
18
Our Algorithm
Image Pre-Processing
1) Multiscale Decomposition
CME Vector Flow Field
2) Spatio-Temporal Filter
3) Non-Maxima Suppression
4) CME Front Characterisation
Kinematics Morphology
19
4) CME Front Characterisation

• Ellipse fit
• Height, Width, Curvature, Orientation

STEREO/Cor1 24-Jan-07
(H.E. Schrank, 1961)
20
SOHO/LASCO CMEs
C2 C3 18-Jan-00
21
SOHO/LASCO CMEs
C2 C3 19-Apr-00
22
SOHO/LASCO CMEs
C2 C3 23-Apr-01
23
STEREO/Cor1 CMEs
SECCHI-A 9-Feb-07
24
STEREO/Cor1 CMEs
SECCHI-A 24-Jan-07
25
STEREO/Cor1 CMEs
SECCHI-B SECCHI-A 24-Jan-07
26
STEREO/Cor1 CMEs
SECCHI-A 24-Jan-07
27
CME Kinematics
LASCO/C2 24-Jan-07
28
Next Steps

• More data distribution of CME kinematics.
• Multiple view points (STEREO) triangulation /
  projection effects.
• Automated front detection space weather
  forecasting.

STEREO illustration
29
Thank You
Acknowledgments

• NRL
• Angelos Vourlidas, Simon Plunkett.
• This work is supported by grants from Science
  Foundation Ireland NASAs Living with a Star
  Program.

jbyrne6_at_gmail.com
30
SOHO/LASCO CMEs
C2 C3 23-Apr-01
31
2) Spatio-Temporal Analysis
Degrees of Freedom Scale, Magnitude Angle
in Space Time
32
Vector Flow Field
Vectors with magnitude and inclination angle
33
Normalizing Radial Graded Filter
remove

• Radially the coronagraph image intensity drops
  off steeply.
• The intensity is normalized by subtracting the
  mean and dividing by the standard deviation.

(Huw et al. 2006)
34
Scale Chaining / Masks
remove
Degrees of Freedom Scale, Magnitude Angle
in Space Time gt Spatio-Temporal Image
Processing Thresholding
35
Method FlowChart
Image preprocessing
Multiscale decomposition
scale chaining (denoising masks)
Gradient space
vector field (angle magnitude)
Spatio-temporal filter (noisy masks)
NRGF radial filter
Combine scale chain spatiotemp gt filter masks
Non-maxima suppression
Ellipse characterization
36
Movie Scripts
Remove

• http//www.maths.tcd.ie/jaydog/Solar/canny_atrous
  /automation/20000118_arrows_combined_rebin.html
• http//www.maths.tcd.ie/jaydog/Solar/canny_atrous
  /automation/20000418_arrows_combined_rebin.html
• http//www.maths.tcd.ie/jaydog/Solar/canny_atrous
  /automation/20040401_arrows_combined_rebin.html
• http//www.maths.tcd.ie/jaydog/Solar/CME_ellipse_
  movies/24jan07/C2_movie_ell.html
• http//www.maths.tcd.ie/jaydog/Solar/CME_ellipse_
  movies/24jan07/C3_movie_ell.html
• www.maths.tcd.ie/jaydog/Solar/STEREO/pb/24jan07/G
  raphs_plots.html