577 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
(a) Before the overlay 
(b) After the overlay 
Figure 2. Images before and after the overlay of FCD and RS image 
3. LOCAL CLUSTER DETECTION FROM FCD 
Local statistics is a hot topic in many fields. Based on Moran’s I 
statistic (Moran, 1950), the local indicators of spatial 
association was derived by Anselin (1995, see also Getis and 
Ord, 1996) to resemble passing a moving window across the 
data, and examining dependence within the chosen region for 
the site on which the window is centered. The specifications for 
the window can vary, using perhaps contiguity or distance at 
some spatial lag from the considered zone or point. With this 
concept, a new statistic is proposed to detect clusters from FCD. 
3.1 Fundamentals 
The local Moran statistic is 
/. = 
n(yj-y) 
Yfy-yf 
i 
'L w tiy J -y) 
a) 
Where, 1 i is the statistic of local Moran’s I at region i , y ■ is 
the attributes of region i , y is the expected value and is 
the weight. Based on the concept of equation (1) and 
considering the character of FCD, a new statistic is defined in 
equation (2) 
L t = max(F') 
F-hzk 
fk 
k . = 'L w «y J 
J 
(2) 
Where, y t is the number of FCD within an area centred at 
point i , Wy is the weight, k is median value of and L t 
is the new local statistic at test time t. 
3.2 Significance test of 
L t Due to the uncertainty of FCD, it is hard to tell what kind of 
distribution L t has. Therefore, Monte Carlo simulations are 
employed to get the critical value. Monte Carlo simulations of 
the null hypothesis of no local clustering confirmed the actual 
value of L t consistent with CC — 0.05 for each time. The 
simulations were carried out by first filling the study area with 
randomly distributed points. The local statistics were then found 
using equation (2). And then the critical value for L t is found. 
Because of making a multiple testing, in which each time L t 
is found to test whether it is significant till no significant L t is 
found, to keep the experimentwise error rate to a specified level 
(usually OC — .05 ), the Bonferroni adjustment (David, 1956) is 
implemented. If k separate and independent tests are made, 
then instead of choosing a critical value of the test statistic 
using CC as the Type I error probability for each test, simply, 
a/k can be used for each test. Therefore, it is an iteration 
process because parameter k of Bonferroni adjustment is 
decided by the times of multiple testing. An iterative procedure 
is needed to approximate the optimal solution. In the case study 
in section 5, the detail procedure will be discussed. 
4. Strategy to construct spatial road network 
4.1 Assumptions