Your SlideShare is downloading. ×
0
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
IGARSS presentation WKLEE.pptx
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

IGARSS presentation WKLEE.pptx

202

Published on

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

  • Be the first to like this

No Downloads
Views
Total Views
202
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
10
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. An Efficient Automatic Geo-registration Technique
    for High Resolution Spaceborne SAR Image Fusion
    IGARSS 2011
    28/July 2011
    Woo-Kyung Lee and A.R. Kim
    Korea Aerospace University
    wklee@kau.ac.kr
  • 2. Motivation
    • As the resolution level improves,
    * the unique feature of the radar imaging becomes prominent and the task of image fusion with optical image becomes complicated,
    * the number of pixels increases and the amount of resources for calculation such as memory and time consumption escalates exponentially.
    To relieve the burden of the work and make it done in real time.
    Efficient image matching in both rural and urban regions
    Simple approach to the SAR image registration and fusion
    Let the machine do the rest of the job
    in almost real time
    One click
  • 3. SAR SensorandGeometricCharacteristic
    • Image processing depends on thesurface characteristics and structures
    • 4. Radar images suffer from unrealistic distortions
    • 5. Non-linear distortions along range, Shortening from shadow region
    • 6. Inaccurate Doppler parameter estimation leads to geocoding errors
    • 7. Unstability in internal system clock and orbit parameters
    System error
    Side-looking Observation
    SAR vs. optics images
    Image acquisition
  • 8. SAR SensorandGeometricCharacteristic
    Effect of Error
    Error Source
    SAR sensor
    - Electronic Time Delay
    - Slant Range Error
    - Incidence Angle Estimation
    - PRF Fluctuation
    Effect of Error
    - Range Location
    - Range Scale
    - Azimuth Scale
    Error correction method
    - Geometric Calibration
    - Deskew
    - Ground Projection
    - Image Rotation
    - Terrain Correction
    Earth
    - Azimuth Skew
    - Range Non-Linearity
    - Foreshortening, Layover,
    Shadowing
    Earth
    - Earth Rotation
    - Side-looking
    - Target Height
    Source of SAR geocoding errors
    Platform
    - Image Orientation Error
    - Squint Angle
    - Doppler Centroid
    Platform
    - Inclination Angle
    - Yaw Angel Error
    - Pitch Angle Error
  • 9. SAR SensorandGeometricCharacteristic
    • Mismatch between SAR and Optical images
    Optics
    SAR
    Geometrical distortions in SAR images
    (a) Azimuth Distortion
    (b) Non linear Range Error
    (c) Deskew
  • 10. SAR Geo-correction with satellite internal data
    • Slant range function has non linear scale variations
    Azimuth
    Slant range image
    Range
    Sland based
    Ground range image
    Ground projection example
    Reference image
    (EO image)
    Ground Based
    • Discrepancy exist compared with the reference image
    • 11. Distortion between SAR and EO are case-sensitive
  • GCP based geometry registration
    • A reference image is chosen to be used for geometric correction or fusion
    • 12. Multiple GCP(Ground Center Point)s are selected and directly applied to individual position error calculation and correction i
    • 13. Based on the selected GCPs, image transforml function is characterized that best describes the discrepancy between the images
    • 14. Original image is re-sampled and re-arranged by the generated transform function
    • 15. To perform geometrical calibration and restore distortion, the GCPs in the SAR images would be re-arranged into the true ground positions
    • 16. It becomes most essential to pick up the best candidates of GCPs
    Basic Principle
    Choice of GCP
    • In general, distinctive features such as road, building, bridges, reflectors are chosen that are easily discriminated for convenience
    • 17. Manually? Or Automatically?? Who will chose what points??
  • Selection of GCPs within SAR images
    • Speckle noise inherence in SAR image makes it difficult to guarantee to pinpoint precise positions that correspond to the reference points.
    • 18. A human work of manual GCP selection is never reliable
    • 19. The number of available points are case-sensitive and still limited by the existence of the distinctive features
    • 20. The precision of the GCP location is not fully guaranteed and the error variance may increase in coarse resolution images.
    Difficulty of GCP choice
    SAR image GCP
    Optical image GCP
  • 21. Methodology
    • Speeded-Up Robust Features (SURF) is known to have highly robust performance
    • 22. Scale, rotation and illumination-invariant feature descriptor.
    • 23. Adaptive for noisy environment and mutll-scale images
    - Only summing operation is involved in producing integral image to match the scale and calculation is speeded up
    • Selection of matching points(GCP) are performed using feature vectors described by Hessian Matrix
    • 24. The size of the constructed Hessian matix can be varied and can be increased to multiple dimensions as desired
    • 25. The number of dimensions is limited by the complexity, time consumption and precision of the image matching.- case sensitive
    • 26. Parameters required for the decision algorithms is set intuitively
    • 27. This work is motivated to find the decision parameters automatically compromising the performance and the complexity
    SURF algorithm
    Selection of GCP and matching
  • 28.                                            
    Block diagram for GCP pair selection
  • 29. Integral image generation
    • For the given image, an integral set of points are summed together
    • 30. The size is variable depending on the scale and complexity of the image
    • 31. Simple summations of intensity levels are performed over two dimensional domain
    : A +B +C + D
  • 32. GCP candidate generation
    • The Hessian matrix , corresponding to each pixel, is simplified by summation with adjacent cells
    • 33. The image scale is varied and the simplified Hessian matrix is obtained for each scale space
    • 34. Harr-wavelet responses are calculated and the feature descriptor is generated
    • 35. The polarity of the image intensity variation is investigated and stored
    Harr wavelet
    X, Y direction
    X direction
    Y direction
  • 36. GCP selection
    • Find a pair of feature descriptors that best fit to each other
    • 37. One-by-one comparison is straightforward but time-consuming and does not guarantee successful matching due to increased ambiguity
    • 38. Construct a look-up table for the feature descriptor
    • 39. Each feature descriptor is indexed depending on their size, variation rate, orientation
    • 40. For a given GCP , a “search process” is performed within other look-up table generated from reference image and the best matching pair is selected
    • 41. Nearest neighbor search is adopted to find the correct matching pair
    Principle
  • 42. GCP pair selection
    - Among the selected GCPs, the Euclidian distance (x, y) – (x’, y’) are estimated to find the nearest points with similar feature descriptor
    • The rates of intensity variations along orientations (denoted as Aand B) are considered as weighting factors
    • 43. Distance multiplication is performed
    • 44. The number of orientation can be increased in order to reduce ambiguity and avoid wrong decision.
    • 45. Appropriate threshold level is required to compare with the distance multiplication and make a decision
    • 46. The GCP match is confirmed when the distance multiplication is less than the threshold level
    • 47. Image Projective Transform function is deduced from the matching GCPs
    Define threshold level
  • 48. Overall procedure diagram for image matching
  • 49. GCP extraction demonstration
    • ScanSAR image is exposed to higher noise level and composed of extended resolution cells.
    • 50. GCP candidates are extracted from both images using the same Hessian matrix structure
    • 51. The number of GCP points appear to be close to each other despite the gap in the image quality
    GCP extraction from SAR images
    (a) Stripmap image
    (b) Scan image
  • 52. Geo-registration demonstration
    • Original SAR image is corrected using GCP matching and transformation
    • 53. Strip mode SAR image over Vancouver, Canada is geo-registered using the reference image in Radarsat-1 SSG format
    • 54. The threshold level is set to be zero for convenience
    881
    557
    Time consumption vs Th. Level
    GCP # vs. Threshold level
    GCP selection for raw image
    Raw
    Reference
    912
    544
    Time consumption vs Th. Level
    GCP # vs. Threshold level
    GCP selection for reference image
    GCP selection
  • 55. Image Matching Demonstration
    • Original image is geo-corrected
    Fusion image
  • 56. Measure of registration errors
    • RMSE(Root Mean Square Error) is calculated for the selected GCPs
    Corrected
    Reference
    The average deviation is about one point pixel size
  • 57. Application to higher resolution images
    • Reference image of TerraSARin EEC format, Toronto,
    • 58. The number of GCP increases consistently when the level of correlation between the two images are high
    As the similarity of the images are high, the GCP increases consistently as the “Threshold Level” decreases
    1.72
    1595
    0.81
    252
    Original
    Reference
    3.21
    2680
    GCP variation rate
    1.47
    404
    GCP selection
  • 59. Image Matching Demonstration
    After fusion
  • 60. Mismatch Error Estimate
    Corrected
    Reference
    The average position error is less than one pixel
    The performance of the matched GCP selection is affected by the image resolution
    Mismatch error is reduced as the image resolution improves
  • 61. Image fusion of SAR over EO
    • This case is where SAR image constitute a small portion of the EO image
    • 62. GCPs from the two images are distinguished
    - The matching GCPs are easily identified by the nearest search algorithm
    (b) LANDSAT EO image
    (a) JERS SAR image
  • 63. Automated threshold level selection
    There exist a turning point, where
    further reduction of the threshold level stops affecting the number of GCP matching points
    Computer traces the variation of the available GCP matches and find the turning point
    • Threshold variance
    Threshold 500, Matching image (14 points)
    Threshold 200, Matching image (15 points)
    Threshold 350, Matching image (15 points)
  • 64. GCP matching and Image transformation
    • Matching GCP selection process stops automatically and image transform function is obtained from the selected points
     15 points extraction
    Feature points extraction
    Find equation
    Transform equation
  • 65. Result of the geo-registration
    Overlaid image
    • RMSE is affected by the resolution discrepancy and inherent image property. - Here it is given as 1.26 pixel
  • Automated Geo-registration software
    • Usually, the threshold level is manually set-up by user looking into the complexity of the images and resultant fusion performance
    • 66. This procedure is replaced by compute search algorithm, where the threshold level is traced to find the turning point
    • 67. Total elapsed time is within several minutes and will be further reduced by adaptive search algorithm
    Original
    Reference
    Corrected
    GCP selection and matching
    Fusion
  • 68. Performance analysis vs. Resolution
    • The number of total GCP is not affected by modes(Stripmap and Scan)
    (a) Stripmap image
    (b) Scan image
    • However, the RMSE is measured as 0.94 for ScanSAR mode while it is 0.34 for stripmap mode
    • 69. It appears that the performance improves as the resolution improves
  • Insufficient information for SAR geometry
    • Internal data within SAR instrument fails to retrieve shadow region
    • 70. There is non-linear discrepancy between slant and ground ranges
    • 71. Generate Errors in geometrical coordinate
    • 72. Need external references to retrieve broken information and correct errors in ground range allocations
    • 73. foreshortening, layover, shadowing
    Limited information
    Shadowing
    Layover
    Foreshortening
  • 74. Impact of the ground characteristics
    • Diverse ground geometry becomes a source of matching errors
    • 75. Mountain areas are severely distorted from the EO case
    • 76. Need to adopt separate transform functions within the image
    After correction
    After correction
    Coast line area
    Mountain area
    Mountain area fusion
    Coast line fusion
  • 77. Matching Performance Comparison
    • Image distortion is not compensated for by the simple GCP transformation
    • 78. Need to divide blocks and adopt modified transform functions separately
    Coast line
    Mountain
  • 79. Modified Transform functions
    • Image is divided into blocks according to the geographical properties
    Average 1.8 RMSE error
    Mountain Area
    Average0.67 RMSE Error
    Urban Area
    Application of separate transform function leads to the reduction of RMSE
  • 80. Summary
    • SURF provides an efficient tool to perform SAR image geo-registration
    • 81. A choice of threshold level is required to perform efficient of GCP matching and it can be automated by tracing its variation curve
    • 82. The image matching algorithm works with various SAR and EO images and the average RMSE is measured to be around 1 pixel.
    • 83. Image blocks containing mountain areas need separate GCP matching and transform function to compensate for image distortion
    • 84. Need to develop optimized transfer function for different type of ground characteristics
    - Indexing of GCP is performed based on their intensity and variation vector
    • With the introduction of adaptive indexing table for selected GCP, the automated image matching is expected to be carried out in real time
    Conclusion
    Further work

×