International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
In order to get this goal, the linear empirical method was used, 
which uses empirical relations between radiant emittance and 
reflectance. This method requires the direct knowledge of the 
spectral signatures of a dark area and a light area, so to 
determine the calibration line. The intersection of this line with 
the axis of abscissas represents the contribution of the 
atmospheric radiance (fig.4). 
To correct the data in thermal bands the PCA (Principal 
Component Analysis) method was used. Preliminary, PCA was 
applied to 10 thermal bands. This method changes the images of 
which DN (Digital Number) doesn't represent value of 
temperature. The most part of information is contained in the 
first principal components. In this particular case the 99.494 of 
the information is contained in the first four principal 
components and the remaining 0.6% in the last six principal 
components. The last six components, containing the noise, 
were deleted instead the remaining four components were 
inverted so temperature values were obtained again. 
5. HIPERSPECTRAL DATA ANALYSIS 
As already said, the presence of underground structures causes 
humidity variations in the ground. These variations affect 
vegetation and some physical parameters such as Thermal 
Inertia and Thermal Conductivity. 
The distribution of vegetation in the first underground can 
underline regular shapes to put in touch with areas with more or 
less humidity. Practically you can notice that vegetation is more 
luxuriant in more often-vegetated ground layers and less 
luxuriant in reduced ground thickness, such as, for instance, 
where there are underground wall structures. The humidity in 
the top underground, due to the well-known capillarity 
phenomenon, moves towards the surface and, during the heating 
up phases, it evaporates, taking away ground and causing a 
temperature decrease. 
The interaction of underground structures with the surface 
characteristics was examined by processing four parameters: 
NDVI (Normalized Difference Vegetation Index), Thermal 
Inertia, Thermal Deviation, Thermal Conductivity. 
S.1. NDVI 
The research of vegetation as an indicator for underground 
structures is based on the spectral response in the visible and in 
the near-infrared (0.8 uum). The more luxuriant vegetation and 
more the reflected energy increases in the near-infrared and the 
reflected one decreases in the red area; this characteristic causes 
that the difference between reflected energy in the infrared and 
reflected energy in the red increases proportionally to the 
vegetation health. 
NDVI was obtained for both images in order to carry out a 
comparison. 
NDVI = (pni PY (Pairt Pr) 
Arctang function was applied to the image obtained in this way: 
this function causes a flattening and an expansion of the image 
histogram, improving the contrast and making the photo 
interpretation easier. 
5.2. Thermal Inertia 
Thermal Inertia is the measurement of the response speed of a 
material to the temperature variations. Giving or taking away 
heat from a object, this object gets warm or cools down quicker 
than another object. Water is characterized by a high thermal 
inertia; this physical characteristic enables to investigate the 
494 
possible presence of underground structures, because, as 
already said, underground structures produce variations on the 
underground humidity. The Thermal Inertia is equal to the 
absorbed heat divided to temperature variations; 
[5 9. 
AT 
To calculate the quantity of heat absorption it is necessary to 
make some premises; the incident energy from the sun is partly 
reflected, partly absorbed and partly transmitted. This behaviour 
is set out in Kirchhoff law 
1=p+T+a 
where | represents 100% of the incident energy of the 
surfaces, 
pz Er indicates the percentage reflected 
Eine 
Ee. ui . 
7 — —— indicates the percentage transmitted 
inc 
  
a= indicates the percentage absorbed 
Eine 
For an opaque body t = 0 thus 
© di Eq 
Izorta«: a=1-p; a=—==1-p; 
Eine 
E. 
Eam) but Enc 
E.sÍt(1-p)-g 
p 
E, is the value of the radiance measured by the sensors, while 
the value of p is extracted from the images calibrated in 
reflectance. 
This calculation was applied to the both images in order to 
generate two images of which DN represents the quantity of 
absorbed energy from the Pixel in the instant of the image's 
acquisition. Successively another image containing the mid 
value of energy absorbed was generated. This image is 
multiplied by the time taken to pass between the acquisitions 
furnishing the quantity of heat absorbed by the pixel. 
The temperature variation AT = T, — T, was calculated 
considering that T5 is the average value of the 10 thermal bands 
of the 12:30 a.m. image and that T, is the average value of the 
10 thermal bands of the 9:30 a.m. image. 
Arctang function was applied to image obtained, containing the 
values of Thermal Inertia. Such a function stretches to a 
horizontal asymptote therefore easing the variability when the 
values of the thermal inertia are raised, giving a better contrast 
of images. 
5.3. Thermal Deviation 
Thermal Deviation is the difference between the local value of 
the temperature and the average value of the surrounding area. 
For this application the image of 9:30 a.m. was chosen because 
especially in the first hours of day, you can notice thermal 
ánomalies due to different evaporation 
AT(xy)-T(xy) -Ta(x,y) 
In 
W