The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1145
4.2.3 Supervised Classification Accuracy Evaluation: Do
supervised classification with the original image and the SFIM
fusion image choosing 5,4,3bands after the bands selection, and
evaluate the accuracy of the classification, the classification
accuracy of the results are showed in table 5
type
XS image
SFIM fusion image
Total
82.81%
90.23%
accuracy
Kappa index
0.7668
0.8673
Table 5.Comparative data of image supervised
classification accuracy
By comparing the total accuracy and Kappa index of the two,
we can see that: the accuracy and Kappa index of the
supervised classification of SFIM fusion image are much
higher than the XS images.
Through the accuracy analysis of the unsupervised
classification and supervised classification, can generally know
the character of the above five algorithm, on the basis of ETM+
classification, select SFIM fusion image as the basic image;
Brovery fusion image as the small information leading to the
lower classification accuracy; HPF fusion image owing to the
better spectral fidelity and more high-frequency information
can be used as the auxiliary image of the visual interpretation;
PCA, ML fusion image has high integration of high frequency
information, which has a certain value in the extraction of the
city internal structure.
5. CONCLUSIONS AND OUTLOOK
This paper introduces the basic concepts and theory of image
fusion, and discussed a variety of fusion method. Summarize
the quantitative evaluation criteria of the mean, standard
deviation, correlation coefficient, entropy, the average grads
and so on, measure and compare each fusion algorithm and
obtained many useful conclusion.
But this study make fusion analysis with the panchromatic
image and multispectral images in the same satellite system of
Landsat-7 ETM+, as different types of sensors have different
data types, it should be more complex in the process of fusion
and the evaluation, yet has to do further studies with specific
issues.
REFERENCES
[1] Zhao Yingshi,2003. Remote sensing application principles
and methods.Bdpng-.Science press,pp..253-254.
[2] Xu Hanqiu. Study on Data Fusion and Classification of
Landsat 7 ETM + Imagery .Journal of Remote
Sensing,2005,9(2), pp. 186-194. 3
[3] Shi Yongjun,2003. Forest remote sensing classification
technology research take the northwest mountain area of
Zhejiang as an example.Zhengjiang: Zhejiang university, 36-
50
[4] Sollerg S.Jain A.K.Taxt.T.A Makov random field model for
classi fication of multisource satellite imagery .IEEE Trans, on
Geosciences and Remote Sensing, 1996,34( 1 ),pp. 100-113.
[5] Wang Wenjie,2006. Image fusion technology based on
wavelet transform.Chinese Academy of Sciences, a master's
degree thesis.
[6] Zheng Yongsheng, Wang Renli,1999. Dynamic monitoring of
remote 5e«5wg-,Beijing:PLA publishing .
[7] Jia Yonghong,2005. Multi-source remote sensing data fusion
technology,Beijing: Mapping Press,.
[8] Smara Y,Belhandj-Aissa A, and Sansal B.Multisources ERS-
1 and Optical Data for Vegetal Cover Assessment and
Monitoring in a semi-arid Region of Algeria [J],International
Journal of Remote Sensing, 1998,19(18), pp. 3551-3568.
[9] Musa M K A,Hussin Y A.Multi-data Fusion for sustainable
Forest Management : A case study from Northern Part of
Selangor ,Malaysia[J]./»/mzatz'o«a/ archives of photogrammetry
and Remote 5e«sz'«g,2000,33(B7),pp.71-80.
[10] Pan Xueqin,2005. The comparison and analysis of data
fusion of remote sensing z'zwage.Kaifeng:Henan university.
Appendix
Supported by the Ministry of Education of Doctor Spot
Foundation (20050147002);
College of Liaoning Province Emphasis Laboratory Item
(20060370)