CS
Mv
Fig.2 shows the relationship between spectral features and
different rating of area with the symptom of progress
disease.
In the spectral region of 640nm, the reflectance decreased
with rating of the damaged area, those were almost whole
area damaged, over 20 percent area damaged, less than 20
percent, less than 5 percent and none by turn.
In the middle infrared region more than 1,400nm, it had
also same order as the visible region.
On the other hand, it had reverse order of the above
mentioned wavelength in the near infrared region of 800
nm.
3. SPECTRAL INDICES AND SYMPTOM OF
DISEASE
3.1 Why and How to Calculate Indices ?
Through the spectral feature measurement, it was identified
to have peak reflectance in the region of 340, 640, 800,
1,450 and 1,940nm.
Rationing is known as one of the spectral enhancement
methods. NDVI(Normalized Difference Ve getation Index)
is the typical index using spectral rationing.
As the effect of rationing, it is possible to reduce some of
influences based on the environmental condition such as
sunlight illumination angle, shadow and etc..
According to the results of spectral measurement, the
relationship between spectral indices and the different
rating of damaged area was discussed. The damaged area
had progress symptom of disease.
As first, in order to enhance spectral information,
following three indices based on the five reflectance peaks
were used:
( R640 - R340 ) / ( R640 4 R340 ) 1)
( R800 - R640 ) / ( R800 + R640) 2)
RidS0 eR 19405 Ale ten 3)
0.9 p
EE A ETT
Equation 1) is the normalized difference of maximum and
minimum reflectances in the region of visible wavelength.
Particularly 340nm had lowest reflectance in the whole
measurement spectral region.
Equation 2) is also the normalized difference by
combining the reflectance in visible red and near infrared
region, it seems to have almost same physical
characteristics of Normalized Difference Vegetation Index.
Equation 3) is the summation of lowest two reflectances in
water absorption regions, those are 1,450 and 1,940nm.
3.2 Indices and Different Rating of Damaged
Area
Fig.3.1 and 3.2 show the relationship between the index
and different rating of damaged area.
Indices' values from 1) increased with increasing of the
damaged area.
The value based on 3) had also same trend as 1).
On the other hand, the value from 2) was identified to have
an inverse trend as the other two indices.
As the results, the differences based on the difference of
rating damaged area were more clear or accurate than the
comparison of reflectance.
Accordingly the above mentioned indices were proved to
be effective for not only spectral information
enhancement but also target feature extraction.
Index Value
oo
QN -
oo
5 Wn
,
e?
-
m -
-
-—7
-
-
see
m h2 UJ
---dÀ--- (R640-R340)(R640«R340)
— 7dà—— — (R800-R640)/(R800--R640)
0 | : à; ;
None less than 5% less than 20% over than 20% Almost Whole
Rating of Damaged Area
Fig.3.1 Relationship between Index and Different Rating Damaged Area
809
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996