International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
The relative mean square root error has the form as 
———— 
ees — Hy)’ 
dm = | ————— 
n- 1 (6) 
where : 4, -B., 
B 
| i 
Hy = 2S 
i =1,2,...,n (n = the number of sample). 
Therefore, the mean square root of single polygon dm is 
>, (4 dm; ) (7) 
“Yin 
dm = 
where single polygon area 4i is the sample weight here, cmi is 
the error of every area unit. 
Since samples are used to replace population, we can consider 
that the dmi is the single polygon mean square root of the 
monitoring region stratum. We suppose that all the polygons 
are not correlative and they are of the same error, the relative 
mean squrae root of the total area of the ith stratum in 
monitoring region dMi is 
_ dm, 
ep (8) 
dM 
  
5. EXAMPLE 
The number of the change polygons is 532 in land Use 
monitoring by remote sensing image in xx city. We sample 173 
polygons with stratified random sampling. The accuracy 
assessment result is given in table 4 and table 5, Using the 
method we suggest in this paper. 
  
  
  
  
  
  
  
  
  
  
  
  
Stratum (1) <10 10-20 20-50 >50 
The number (ni) 72 3 43 28 
The total of 414 484 1460 4142 
Area (Si) 
Relative error(dVi) | 4.51% 1.76% 2.76% 5.90% 
A V(dlvi) 10.4% 12.9% 15.3% 12.1% 
MSR (dmi) 35.4% 16.3% 8.1% 13.0% 
MSR of single 14.9% 
polygon(dm) 
  
  
Tabled. The accuracy assessment of single polygon. 
  
  
  
  
  
  
  
  
Stratum(i) <10 10-20 20-50 250 
The number (ni) 239 103 89 39 
The total of area (Si) 1233 1378 2793 8097 
RMSR (dmi) 35.4% 16.3% 8.1% 13.0% 
RMSR of 2.2% 1.5% 0.8% 1.7% 
total area (dMi) 
  
MSR of total area 
/relative error 
  
  
233 / 1.69 
  
  
Table5. The accuracy assessment of the wholregion. 
6. CONCLUSIONS 
This techniques has been tested in Land use Dynamic Detection 
project conducted by the Ministry of Land and Resources 
P.R.China in the latest years.But because of the complexity of 
accuracy assessment, we should consider many factors, such as 
the hardness of field collection ,the expending of time and 
money,etc. The implement of accuracy assessment should 
depend on the cooperation of all departments and the 
improvement of accuracy assessment techniques. 
1. Siamak Khorram, 1999. Accuracy Assessment of Romote 
REFERENCES 
Sensing Derived Change Detection. ASPRS. pp. 131-135. 
2. Rosenfield, G.H., 1982. Sample Design for Estimating 
Change in Land Use and Land Cover. PE&RS, Vol.48, No.5, 
pp.793-801. 
3. Stephen V. Stehman and Raymond L.Czaplewski , 1998. 
Design and Analysis for Thematic Map Accuracy Assessment : 
of Environment 
Fundamental 
Priniples, 
Vol.64, No.3, pp .331-344. 
4. Zhou 
Zhejiang,China. 
458 
Remote 
Sensing 
Sheng,etc,1988.Probability and Statistical. 
KEY 
ABS 
Auto 
data 
detec 
detai 
betw 
detec 
GIS 
diffic 
unch: 
algor 
featu 
re-gr 
With 
netwi 
parts 
most 
key | 
devel 
netwi 
manu 
meth 
extra. 
(PIC 
featui 
detec 
litera 
For I 
impoi 
short 
integi 
Mark 
(Dai, 
(Dres 
GIS | 
inclu 
tradit 
[i (2i 
simul 
chang 
have 
modu 
in th 
relati 
literat 
from 
algori 
the of