Accuracy
type"v
Producer’s
accuracy
User’s
accuracy
Overall
accuracy
cropland
88.24%
86.71%
vegetable fields
84.00%
87.50%
86.42%
housing estate
81.02%
81.00%
Table 2. Accuracy evaluation by confusion matrix
for image classified by improved fuzzy c-mean classifier
The same accuracy evaluation was finished for the image
classified by ERDAS, the overall accuracy is 79.2%.The test
shows that the fuzzy c-mean classifier is a high precision
method for image classification.
5 CONCLUSION
In order to solve the remote sensing image problem of
uncertainty and mixing pixel, improved fuzzy c-mean
classifier of soft classification method which adopted
Mahalanobis distance was introduced in this paper.
Mahalanobis distance scale suits to the real
conditional of pixels distributed can detect
characteristic space for data muster of super ellipse structure.
The using of it is advancement relative to the fuzzy c-mean
classifier which use the Euclidean distance.
After interpreted the theory of improved fuzzy c-mean
classifier, improved fuzzy c-mean classifier was constructed
to classify a image, at last Confusion matrix was introduced to
assess the accuracy of classified image, then comparation was
made between the soft classified image and the one classified
by ERDAS, the test proved that the land cover classification
by improved fuzzy c-mean classifier has very high accuracy.
The method also have disadvantages, it relies on the original
clustering center awfully because the aim function is
degressive during iteration. So if the original center departs
from the idea one, the fuzzy c-mean method will get into a
minimum answer locally, but the whole optimum answer.
Genetic Algorithm which is overall optimum method can
be applied in classifying in the future to avoid that problem.
Besides, it has large calculation amount and can spend much
time, all those problem should be studied ulteriorly.
3. Huang Ning, Zhu Minhui, Zhang hourong,2002.
Considering neighborhood information in image fuzzy
clustering..Journal of electronics,July,pp.56-59.
4. Friedl M A,Brodeley C E. Decision Tree Classification of
Land Cover from Remotely Sensed Data [J], Remote Sense
Environ. 1997,(61):399-409.
5. Gao Xinbo,Pei Jihong,Xie Weixin,2000.study of weighted
index m in fuzzy c-mean algorithm. ACTA ELECTRONIC A
SINICA,Vol.28,April pp.98-102.
6. Hasbaga, Ma Jianwen,Li Qiqing, Liu Zhili,Han
Xiuzhen,2004. fuzzy c mean improving and comparing the
clustering of satellite image.Computer Engineering, vol.30
Nol 1,June,pp.37-41.
7. Foschi P G,Smith D K,1997. Detecting Sub-pixel Woody
Vegetation in Digital Imagery Using Two Artificial
intelligence Approaches. Photogrammetric Engineering &
Remote Sensing,63(5):493-500.
8. P. Gong, J. D. Marceau, and P. J. Howarth,1992. A
comparison of spatial feature extraction algorithms for
land-use classification with SPOTHRD data. Remote Sensing
Environment,40:13 7-151.
ACKNOWLEDGEMENTS
Our research project was supported by the ’’Excellent Youth
Research Fund Program of LiaoNing Technical University
(N0.O6AOI) ”, and the ’’Open Research Fund Program of the
Geomatics and Applications Laboratory of LiaoNing
Technical University (No. 2006008) ”.
REFERENCES
1. Wang Juan, 2007.Study about the Recognition Method of
Land Cover with Remote Sensing Image Based on Fractal
Texture. Liaoning Technical University,FuXin.
2. I. Nedeljkovic,2004.image classification based on fuzzy
logic . The International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Sciences, Vol. 34,
PP.121-130.
