167 
samples around that sample are selected and are judged by the 
similarity between the samples of Tl. 
Firstly, the change index was judged by triple standard 
deviation, if less than the 3 times of deviation, then mark the 
land use segment, vice versa. 
Secondly, each segments of T2 will be classified according to 
the minimum Euclid distance to the T2 sample segments 
accepted, and the corresponding land use type will be assigned 
to the current segment. 
II |2 
d , =F-^j| = JZK ~ m v\ (5> 
If d i — min {d k } , then the corresponding land use type k 
will be assigned to the current segment. 
Figure 2. quick bird image conflation with the land use map 
3. EXPERIMENTS AND DISCUSSION 
3.2 Pre-procession 
Based on the new methodology, we do experiment by means of 
quick bird image data of 2002 and the land use map of 2002, 
the main steps are given here: 
]-i 
registration f- Land — registration 
1 map 2002 1 
K 
SamDle seament 
■ T 
I® 
Texture 
character 
Feature 
extraction 
. 1 
I ! 
generation 
(sample) 
- K 
a. 
JLL 
Sample 
judgment 
Submission result 
feature 
generation 
(Segment) 
▼ 
Change 
detection 
1 
Figure 1. the flow chart of LUCC based on the land use 
segment 
3.1 Main datum 
1:10000 land use map 
1:5000 wuhan map 
1:10000 wuhan DRG 
Quick bird satellite data in 2002, 2005 
3.2.1 Registration 
In experiment, we pick 35 GCPs and 6 check points, make use 
of 3 order polynomial to geometric correction , projection: 
Transverse Mercator, Spheroid: Krasovsky, Datum: 
Krasovsky, total control point error is 0.2376, less then 0.5 
pixel. 
3.2.2 Standardization 
In order to reduce the diversity of spectra in different time 
series image, we standardize the image in 2002 and 2005. 
Figure 3. standardization of quick bird image 2002 (band 5)