Generated Reference Ground Control Points from SAR Image
sequential similarity detection algorithms technique (SSDA)
After the transformation of GCP, the obtained result image in te previous section and OPS image as shown in figure 7 will be divided in order to determine the specific area such as road intersection, river intersection. The subimage will pass to the registration process. It is and inherently basic requirement for any image processing system. Te SSDAs in this paper are efficiently digital implementation for determining correspondent prints translation will yield the required registration . this paper does not concern about the problem of arising form magnification and rotation . however, the method are applicable for the proper modification.
Figure.8 Illustration of Search Space
Let two subimage, S the search area, and W the window is defined as shown in figure 8. the usual normalized correlation is defined as follow
[2].
The reference image from SAR image enhancement consists of the details concerning about road and river ( canal ) which are used to generate GCPs as the OPS ( Optical Sensor ) image should be corrected. There are fog and the oher information appearing with road and river, thus the SSDA technique by using normalized correlation (2) is quite difficult and the high misregistratonn is possible. This paper proposes to apply the static correlation measurement. The die correlation by convolution the OPS subimage with whitening filter [5] as follow.
| r2 | -r(1+r2)
| r2
|
| -r(1+r2) | (1+r2)2
| -r(1+r2) |
| r2 | -r(1+r2) | r2 |
|
r=0
|
r=1.0
|
r is the adjacent element correlation and 0<
r<1 . for
r=0, the correlation is normally and no decorrelate. AS
r is increased, especially
r=1, the correlatin occurs as edge detection and the whitening filter is in the gradient operation
Fig.9 Decorrelated result by convolution with whitening filter, when r=0 and 1, respectively.
(a) Correlation result r=0
(b) Correlation result r=1.0
Fig. 10 Results from static correlation measurement
Experimental Results
The static correlating measurement for specific area is used by using he sub image of OPS and SAR images which has the peak opposition and assigning to be GCP of image. Having used this method by applying with the road intersection and river ( canal ) intersection will obtain the enough GCP points for doing the geometric correction but it depends on the selected spatial transformation method . Such as the bilinear interpolating uses 10 points as the cubic spine interpolation uses 16 points. Figure 10 shows the image which is generated by GCP form SAR image. In this case will show only 4 point, but the correction depends on the quality of OPS sub images. The quality means the percentage coverage area of cloud, fog .
Fig .11 Generated GCPs on OPS by SAR image
If there are a lot of cloud, the generate GCP may not by used. As the small cloud or some part, the generate GCP may be easily.
References
-
R.Bernstein and H.Silverman , "Digital techniques for earth resource image data processing ," Aeronautics and Astronautics 8th Annu. Metting, Vol C21, AIAA paper no. 71-978, oct .1971.
-
P.M. Bakis and R.Wesley, ERTS Program : Final report on an all -digital approach to image processing for ERTS, IBM Research Center, Yorktown Heights , New York , March 1970.
-
P.M.Bakis and R.Wesley, " Digital images processing for the earth resources technology satellite data," ASP Meeting, Washington D.C. , March 1971.
-
R.C.Gonzalez and R.E.Woods, Digital images processing, Addison-Wesley Publishing Company, USA , sep. 1993.
-
W.F. Webber , " Techniques for image registration ," IEEE Conf. on Machine Processing of Remotely Sensed Data , pp. 1B.1-1B.7, Oct. 16-17, 1973.
