ue 
esent 
as 
putation 
mates 
istort 
image 
wn 
45°, 
this 
in tex- 
same 
XHH 
o flight 
711). 
pproach 
on of 
of the 
rn in a 
r mo- 
ons. 
no a 
on of 
The 
n value 
AR 
shape, 
idual 
e- 
ege- 
S- 
cit is 
pe as 
n the 
m the 
Ops 
n 
Although the spatial resolution element of the SAR 580 is substantially larger 
than the row spacings in wheat, barley, corn, or sugar beet fields, certain 
within-field textural variations are evident in the image (see Fig.5) and this 
is exactly the point of interest. By developing a digital measure of local 
image texture which was rotationally invariant, I hoped to be less dependent 
on illuminating, geometry than the raw SAR backscatter data. I also was seeking 
a classification procedure which could be applied simultaneously to large fields 
(in my case 200 m x 600 m) and to narrow fields (10 m x 200 m) without having 
to manually delineate the field boundaries. 
Statistics for representative areas of certain crop types are given in Table 1. 
Mean values and standard deviations were computed both in the raw data and in 
the average inertia image. The mean values in the inertia image were distri- 
buted over a broader range than in the raw image. Unfortunately, their standard 
deviations increased also. To overcome this I applied isotropic filtering with 
a radius of 2 pixels to the inertia image. The result of doing this to the 
image is shown in Figure 7. 
  
Table 1. Statistics of Representative 
Crops in Data Set 709 (Xp) 
Raw Data "Inertia" 
Std. Std. 
Crop Mean Dev. Mean Dev. 
Wheat 1 24.54 4.20 36.03 7.60 
Wheat 2 28.79 5.31 44.26 9.17 
Barley 1 35.47 7.78 66.47 13.48 
Barley 2 33.49 6.55 64.36 11.42 
Grass 1 45.22 8.95 110.81: :19.71 
Grass 2 46.07 10.28 128.94 26.40 
Corn 1 57.02 12.23 185.16 . 34.60 
Corn 2 51.67 10.75 154.39 23.49 
Sugar Beets 1 60.30 13.55 198.86 27.26 
Sugar Beets 2 72.13 15.82 236.183 25,04 
Trees 1 49.06 16.07 211.11 . 34.68 
Trees 2 59.32 18.44 39.65 
226.92 
237 
wr 
 —B I eh NH JE A SST rt LE ER