stanbul 2004
sed on the
, we adopt
s function;
iass function
threshold #4,
ies by each
h.
than
ler than
lcation can't
cation based
en as follows:
‘each feature
bining rule,
nd its Belief;
es to choose
N
| some aerial
sification is
area, water
» from each
100 by 100
ice of multi-
"s evidential
are utilized,
are (fractal
1d the fused
h the results
tained from
lassification,
opy of gray
ting texture.
e based on
sification is
and
International Archives of the Photogrammetry, Remote Sensin g and Spatial Information Sciences, Vol XXX V, Part B3. Istanbul 2004
Confidence interval [Belief(A), Plausibility(A)] Identified
Feature |m(0) ; : =
Texture Woodland Grassland |Inhabitantarea| Water field result
Fractional|0.482| 0.280 | 0.762 | 0.089 0.571 | 0.050 0.532 0.110 0.592 Unknown
Woodland Entropy 0.162| 0.678 | 0.840 | 0.087 0.249 | 0.032 0.194 0.051 0.212 Woodland
Fused |0.105| 0.727 | 0.832 | 0.081 0.187 | 0.030 0.136 0.060 0.165 Woodland
Fractional|0.446| 0.132 | 0.578 | 0.225
0.671 | 0.119 | 0.565 0.177 0.623 Unknown
Grassland Entropy 0.233} 0.115 | 0.343% | 0355
0.588 | 0.072 0.305 0.164 0.397 Unknown
Fused [0.162 0.121 | 0.283 | 0.481
0.643 | 0.066 | 0228 | 0.170 | 0.332 Grassland
Inhabitant |Fractional|0.543| 0.137 | 0.680 | 0.126
0.669 | 0.283 | 0.826 0.101 0.644 Unknown
area Entropy 0.475! 0.062 | 0.527 | 0.089
0.564 | 0.263 | 0.728 0.140 0.615 Unknown
Fused |0.234| 0.080 | 0.314 | 0.137
0.371 | 0.341 | 0.635 | 0.154 0.388
Inhabitant
area
Water field|Fractional|0.426| 0.145 | 0.571 | 0.181
0.607 | 0.119 | 0.545 0.238 0.664 Unknown
Entropy 0.1621 0.113 | 0305 | 0183
0.375 | 0.148 | 0.340 | 0.464 0.656
Water field
Fused |0.182| 0.095 | 0.277 | 0.171
0.353 | 0.110 | 0.292 | 0.463 0.645
Water field
Tabel 2. Part of confidential intervals and uncertainty probability
Table 2 is part of confidence intervals and uncertainty
probabilities. The confidence interval obtained from the fused
feature is smaller than that from the single feature
correspondingly, Belief and Plausibility obtained from the fused
feature are higher than that from the single feature
correspondingly, the multi-feature fusion technique based on
Dempster-Shafer's evidential reasoning for classification is apt
to identify textures correctly. Comparison with uncertainty
probability m(6), the feature fusion technique reduces m(0) and
enhances the power of identify.
4. CONCLUSION
A new multi-feature fusion technique based on Dempster-
Shafer's evidential reasoning for classification of image texture
is presented. The proposed technique is divided into three main
steps. An example is provided. The performance of the method
is investigated with some aerial photos in some area. Compared
with the results obtained from the single feature, the results
obtained from the multi-feature fusion indicate the multi-feature
fusion technique based on Dempster-Shafer's evidential
reasoning for classification is stable and reliable, and efficiently
improve the accuracy of classification.
S.REFERENCES
Galloway M., 1975, * Texture classification using gray level
run length", Computer graphics and Image Processing, 4,
ppl72-179.
Haralick R. 1979, Statistical and structural approaches to
texture. Proceeding. IEEE, 67(5), pp786-804.
Huang Guilan, Zheng Zhaobao, 1998a,The application of
fractal geometry in image texture classification. Journal of
Survey& Mapping, 24(1), pp283-292.
Huang Guilan, Zheng Zhaobao, 1998a, Texture models applied
in image texture classification. Journal of WTUSM, Vol.23,
No.1, pp40-43.
Li Deren, Zhanglixian, 1993a,The statue and methods of image
texture analisys. S&T of WTUSM, No.3, pp30-37.
665
Li Deren, ZhangJixian, 1993b,The statue and methods of image
texture analisys. S&T of WTUSM, No.4, pp16-25.
P.L. Rogler, 1987,Shafer-Dempster reasoning with application
to multisensor target identification systems, IEEE Trans, on
S.M.C, Vol.17, No.6, pp968-977.
Yang Jingyu, Wu Yongge, Liu Leijian Etc, 1994, Data Fusion
Technique in war field. Weapon Industry! Beijing , pp66-67.
