International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
be found in the concrete anomalies. 
Fig. 6 shows AT of the concrete including anomaly sites and the 
distribution of the concrete radiant temperature of these pixels 
is shown in Fig. 7. Four shadow conditions mentioned above 
were also sampled from the images. For each shadow condition, 
three lines of five pixel long were taken based on the inspection 
report in section 4. The five-pixel lines were sampled as 
follows: The points where the concrete structure problem were 
found in the check activities were sampled as the center pixel 
(pixel 0) and the four neighboring pixels (-2, -1, 1, 2) were 
sampled along the road direction (not along the slope aspect, for 
some pixels can be covered with vegetations). Each profile in 
Fig. 6 was taken from the averages of three lines. 
AT under SM and that under SB are almost the same as those of 
normal concrete. On the other hand, AT under SA is larger and 
that under SN is smaller than those of normal concrete. These 
increase and decrease can be considered due to anomalies. It is 
likely that the concrete with anomalies keeps its temperature 
stable once it reaches a certain degree. Judging from the wide 
distribution of TEA, the sampled pixels under no shadow may 
partly contain vegetation or other land cover types. 
11.0 
  
x x 
doas. en. x x 
9.0 
7.0 
= 30 
3.0 .- — em Mei Br -- 
1.0 
vr 
-1.0 -2 -1 0 l 2 
Pixel 
——— : NS ee SM AA ASA SB 
Fig.6 Distribution of AT for Concrete Including Anomalies 
30.0 
20.0 + 
+ 000 
. * 
o 
= + 
= + 
ë + 
à. + 
be 
10.0 * 
eoo M, M 
0.0 
0.0 10.0 20.0 30.0 
TEA(?C) 
€: NS m: SM a: SA x: SB 
Fig. 7 Radiant Temperature Distribution of the Concrete 
Including Anomalies 
590 
The temperature distribution under SA and that under SB are 
also similar to those of normal concrete. It seems hard to detect 
the concrete anomaly under those shadow conditions from the 
temperature distribution. The temperature distribution range of 
the concrete with anomaly under NS is wider than that of 
normal concrete. It might be inferred from this distribution 
range that there are several types of concrete anomaly which 
can be detected by the thermal data. The distribution range of 
Tea under SM is wider than that of normal concrete. It is not 
hard to imagine that the heat capacity of the concrete with 
anomaly is different from that of the normal one. 
For a statistical interpretation of concrete anomaly, the T-test is 
proposed to detect the temperature anomaly at each pixel. 
T i / 
Z snp = Xstnp-Misı/ Ost (2) 
where Z : Statistical score 
X : Radiant temperature of the pixel 
M : Mean value of the radiant temperature of normal 
concrete (Shown in Tab.3) 
o : Standard deviation (STD) of the radiant temperature 
of normal concrete (Shown in Tab.3) 
s : Shadow condition (N : NS, M: SM, A: SA, B: SB) 
t : Observation time (A : EA, M : EM) 
n : Number of line (1, 2, 3) 
p : Pixel position (-2, -1, 0, 1, 2) 
If |Zsanpl Of |ZsMnpl is greater than 1.96 (95% confidence 
interval), the pixel will be-discriminated as a concrete anomaly 
point. 
  
Table. 5 shows the results of T-test for the concrete anomaly 
points and the neighbor pixels. It can be seen that the irregular 
temperature is found not only at the pixel locations sampled as 
concrete anomaly pixel based on the report but also its neighbor 
pixels. 11 of 12 center pixels (Pixel 0) in the sampled lines were 
discriminated as concrete anomaly points. This accuracy 
presents that T-test is very appropriate for detecting the concrete 
anomaly. However, the results of the discrimination with (2) 
Table. 5 Result of T-test for the Concrete Anomaly Points and 
the Neighbor Pixels 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Shadow Condition |Line £| Time um = 
I EA 12.54]. -2.38 | .-2.08 
EM 22.951 4.26 | 421 
NS 2 EA [22r] -180] -1.74 
EM 2423| 4.88 | 4.75 
3 EA -3,09 | -2.67 | -3.07 
EM -0.92 | -0:41 0.23 
| EA (2.40 |. -2.87 |. -4.16 
EM 740.63 0.74 0.32 
SM 2 EA -2.02 1.51 0.09 
EM 047} -2.21 | -0.95 
3 EA -9:98 | -10.71 | -10.89 
EM 1:26 1.58 1:95 
I EA 2.55 8.10 2.35 
EM 0.24 | -0.06 | -1.02 | -0.59 | -0,41 
SA 2 EA -0:05-| -0.60 | 1-10 | -0.8$ | 0.00 
EM -0.13 | -0.18 | 0.04 | -045 | -0.58 
3 EA -3.25 | -2.40 | -2.65]| -1.50]| -1.40 
EM „1611-1821 -L511 -1.51] -1.52 
; EA -039] 021] 052] 0.11] -0.71 
EM 3.24 | 2.41] 2.06| 1.94| 0.47 
SB 5 EA L16] TAL] LL 1.52 [22:46 
EM 1.00 |- 2.76 | 3.12] 1.35 [28:76 
3 EA 136| 154| E54] 143] 102 
EM 482|] 1.65] 3.18] 3.00] 3.71 
EE : Pattern. 1 : |Zspnpl >= 1.96 and |Zsmnpl >= 1.96 
   
: Pattern.2 : |Zspnpl >= 1.96 and |Zsmnpl < 1.96 
: Pattern.3 : |Zsbnp| < 1.96 and |Zemnp| ># 1.96 
Intern 
depen 
of the 
The a 
patteri 
are di 
mostly 
imply 
clue fi 
anome 
No pi: 
under 
20fS 
are 180 
cover | 
Actual 
becaus 
distrib 
sugges 
detecti 
In this 
remote 
detecti 
measu 
anoma 
Statisti 
anoma 
remote 
analysi 
are gre 
Future 
DTM | 
multite 
fusion 
inform 
detecti