A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR SERIOUSLY OBLIQUE AERO IMAGE

1
A SELF-ADJUSTIVE GEOMETRIC CORRECTION METHOD FOR
SERIOUSLY OBLIQUE AERO IMAGE

• IGARSS 2011 Vancouver, 24-29 July

Chunyuan Wang, Ye Zhang, Pigang Liu, Qi Xu,
Yanfeng Gu from Harbin Institute of Technology,
China
2
Content
Motivation
Analysis of the Projection Errors
Method System
Experiments Results
Conclusion
3
Motivation

• Geometric Correction
• Importance
• Important preprocessing of remote sensing image
  processing and applications
• Special Situation
• Image taken in a large angle
• Two Important Problems
• Projection errors caused by
• curvature of the earth relief

4
Content
Motivation
Analysis of the Projection Errors
Method System
Experiments Results
Conclusions
5
Analysis of the Projection Errors

• Projection errors caused by relief
• Linear displacement between image points
• aa image point displacement caused by relief
• f the focal length of the sensor
• h the relief height
• H imaging height.

6
Analysis of the Projection Errors

• Projection errors caused by curvature of the
  earth

Projection errors caused by the curvature of
the earth increase with the increasing view
angles
7
Content
Starting point
Analysis of the Projection Errors
Method System
Experiments Results
Conclusions
8
Method System

• Polynomial correction model
• A fitting method using control points.
• The quadratic term effective correct projection
  errors caused
• by the
  curvature of the earth
• The third dimension effective correct projection
  errors caused
• by relief via
  Digital Elevation Model
• Ternary quadratic polynomial

(for i,j,k0,1,2)
9
Method System

• In practice, only depending on ternary quadratic
  polynomial to
• correct projection errors caused by both
  curvature of the earth
• and relief, the correction error is relatively
  large.
• Relief-projection error is more complex with a
  curvature terrain.
• The adjustable ternary quadratic polynomial
• Based on the generated characteristics of
  projection errors,
• suppose y direction is the direction of large
  view angle, the
• improved polynomial model is

(for i,j,k0,1,2)
10
Method System

• Polynomial correction process with
  self-adjustable model

11
Content
Starting point
Analysis of the Projection Errors
Method System
Experiments Results
Conclusions
12
Experiment

• Dataset gather from our simulation imaging
  system
• Imaging in the curvature surface on our earth
  model with
• large view angles.
• The control points and test points in all the
  experiments have
• the same quantity and quality.
• Two criterions
• Root mean square error (RMSE)
  correction accuracy.
• Location errors of high objects (LER)
• recovery accuracy of the roof location.

13
Experiment 1

• Correction of curvature -projection error

80 degrees distorted image Quadratic
polynomial Affine correction The accuracy of
correction (/pixels)
14
Experiment 2

• 2. Correction based on the self-adjustable model

Reference image Ternary quadratic
polynomial
65distorted image
Ternary cubic polynomial Adjustable model
15
Experiment 2

• 2. Correction based on the self-adjustable model

16
Conclusion

• For correcting the seriously distorted
  image, the new self-adjustable ternary quadratic
  polynomial model alleviates the seriously
  distortions problem caused by relief and earth
  curvature and recovers the height objects
  location better. It is experimentally
  demonstrated that self-adjustable polynomial
  model outperforms the conventional models and is
  effective for the seriously distorted image
  acquired in large view angles.

17
Thank You

• Email cyuanw_at_163.com