You are on page 1of 5

Module 8 Slant range to Ground range

conversion

Program to run: sr2gr.m (bilinear interpolation),


srogr_cubic_conv_2d.m (cubic convolution interpolation)
Functions used : -

Input : \module_7\multilooked_azi.dat

Output : sr2gr.dat (bilinear interpolation), srogr_cc.dat (cubic convolution interpolation)

Description :
The SLC and multi-looked images have slant range geometry i.e. it is based on the actual
distance between target and radar (in a direction which is at an angle/slant to the ground) not the
distance measured along the earths surface. We need to convert the slant range measurement
into the true ground range measurement so that all the features are in their proper planimetric
(x,y) position relative to one another in the final radar image. This is done by resampling in range
direction. In the program given in the module, sr2gr.m, both range and azimuth pixels are
resized and resampled to get aground range image. Run sr2gr.m to perform slant range to
ground range image conversion on SLC or azimuth multi-looked image along with bilinear
interpolation. Run srogr_cubic_conv_2d.m to perform slant range to ground range conversion
and interpolation by cubic convolution method.

Explanation :

Slant range to ground range conversion:

Slant range to ground range conversion is also done in patch-wise manner. Azimuth
resampling: azimuth resampling is done by forming indices in azimuth direction using the
following relation.
az lookup (r index )=new r /(line spacemult lk )
index

Where r_index is the corresponding existing index, az_lookup is the new index array that
extends from zero to number of azimuth lines in one patch. The new_size specified can be
changed to vary the size in azimuth direction and also decides the size of az_lookup.

r g * Re

r g / Re

1 / 2
cos 2 sin 2
Radius of the Earth Re
a2 b2

Range resampling: using the cosine law, the ground range can be determined by using the
relation
htht + roro
rg=acos ( 2ht )
Where ht is distance of sensor from the center of the earth, re is the radius of the earth and ro is
that slant range distance. This is calculated for the initial slant range. From this value of rg, the
new_size required is added and slant range and corresponding ground range are iteratively
calculated. Slant range calculation for a given ground range is done using the following relation

(
rp= htht+ 2.0htcos ( ))
rg

The new range lookup index is created iteratively using the relation
gr lookup (c index )=(rpro)1000/ pixelspace

After this resampling is completed, interpolation is done. Interpolation can be done in


several ways like bilinear interpolation, cubic convolution interpolation, cubic spline
interpolation etc. Bilinear interpolation is used for resampling the input slant range image to get
ground range image. Using bilinear interpolation, the output pixel value can be written as
4
zk

P i=

k=1 ( )
D2k
4
1

k=1 ( )
D2k

Where zk is the value of the neighboring pixel and D k is the distance to that pixel. This results
in weighting of the pixels according to the distance from the resampled point. The output pixel
value will depend on four neighboring pixels.

Cubic Convolution interpolation: cubic convolution is a polynomial interpolation method


based on convolution. The weighting is done in 4 x 4 window in contrary to the 2 x 2 in bilinear
interpolation. The third order cubic convolution kernel is given below.

3 3 5 2
|x| |x| +1 if 0 x 1
2 2
1 3 5 2
|x| + |x| 4|x|+2 if 1 x 2
2 2
0if 2x
cc 3=

This is a one dimensional windowing function. The 2-Dimensional cubic convolution is


achieved by applying this window in two dimensions. To determine each pixel value, weighting
is applied for each of the 4 rows and then the same window is applied to the 4 values obtained.
Results:
Ground Range Image

100

200

300

400
Azimuth bin

500

600

700

800

900

1000
200 400 600 800 1000
Range bin

Ground range image for new size in range and azimuth direction - 20
Ground Range Image of Mumbai

100

200

300

400

500

600

700

800

900

1000

1100

200 400 600 800 1000

Ground range image after applying cubic convolution interpolation

You might also like