|
|
|
Poster Session 1
|
Recognition of Flooded Area in Radar Image using Texture Feature Analysis
3 Results
3.1 Extraction of Flooded Area
The basic method for water detection is thresholding. A number of threshold levels can be defined to separate various ranges of texture value. In this research the segmentation was performed on the basic of the characteristics of the double peak in the histogram of texture images as shown in figure 1. We choose the value located at trough point as the threshold. Figure 2a,b,c respectively represent the flooded area from the above texture images. It was easily found in the result images that the areas shadowed by mountain were mistakenly detected as flooded area. By using the DEM these areas can be automatically detected from the derived images.
Compared with the ground truth , an image interpreted visually from SAR data (shown as the contour line of water bodies), We can find that the main errors distribute in ramification. Of which the result of the extracted water segments using homogeneity feature was best, the pixel number was 86990 version 74811, the accuracy was about 86%. The method of entropy features was better the accuracy was 80%. The other accuracy which extracted water variable feature was 72%.
3.2 land Use Statistics in flooded Area
Overlying the map of country's boundary with the land use map interpreted from landsat TM and the flooded area image, flooded land use was calculated and tabled in the following table. In this flood, the most affected was the paddy field, second grass land (shown as figure 3).
Table 1 Area of land use in inundated region of Poyang lake
| Country Name |
paddy |
Non-irrigated |
forest |
grass |
gloodplain |
| Huangei |
3379.460 |
195.860 |
3.600 |
16.890 |
1180.200 |
| Jiujiang |
657.960 |
1230.340 |
32.760 |
727.630 |
843.456 |
| Pengze |
19.701 |
108.007 |
33.346 |
3.960 |
78.750 |
| Yongxiu |
5639.240 |
9.391 |
0.000 |
67.391 |
18.279 |
| Huku |
996.165 |
676.772 |
206.913 |
1771.110 |
2035.360 |
| Ruichang |
2462.630 |
0.000 |
0.000 |
0.000 |
0.000 |
| Jiujiang City |
699.119 |
16.216 |
76.590 |
850.526 |
1438.220 |
| Duchang |
7478.880 |
3084.700 |
1068.750 |
17616.800 |
5150.540 |
| Xinzi |
10639.800 |
1110.850 |
1572.070 |
7656.030 |
2023.320 |
| Xinjian |
24639.800 |
1412.850 |
2217.230 |
11406.000 |
3110.130 |
| Jiujiangi |
3344.590 |
92.250 |
0.000 |
17.920 |
81.878 |
| Total |
60310.05 |
9736.486 |
5211.259 |
40164.257 |
16560.133 |
Reference:
-
Arai K., 1991, Multi-Temporal Texture Analysis in TM Classification. Canadian Journal of Remote Sensing, Vol. 17 No. 3 : 263-270.
- Barber D.G. and Ledew E.F., 1991, SAR sea ice discrimination using texture statistics A multivarice approach, PE&RS, 57(4): 385~295.
- Conners R.W. and Harlow C.A., 1980, A Theoretical Comparison of Texture Alogrithms, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(3): 204~222.
- Haralick R.M., 1979, Statistical and Structural Approaches to Texture, Proceedings of IEEE, 67(5) 786~804.
- Laws K., 1985, Goal Directed Texture-Image Segmentation. SPIE-Application of Aritifical Intelligence, Vol. 548: 19-26.
- Lee J.H. and Philpot W.D., 1991, Spectral Texture Matching: A classifier for Digital Imargery. IEEE Transaction on Geoscience and Remote Sensing, Vol. 29 No. 4:545-554.
- Sun C. and Wee W.G., 1982, Neighboring Grey Level Dependence Matrix for Texture Classification Computer Vision, Graphics and Image Processing, Vol. 23 :341-352.
- Sun Y., Carlstron A. and Askne J., 1992 , SAR Image Classification of Ice in the Gulf of Bothnia. Int. J. Remote Sensing, 13(13): 2489~2514.
- Wilson P.A., 1997, Rule-Based Classification of water in Landsat MSS Images Using the Zhou Chenghu, Study to information system of flood disaster estimation, 1993, Beijing.
Science and technology Published house.
Figure 1: Histogram of texture images
Figure 2:a The extracted water segments using homogeneity feater
Figure 2:b The extracted water segments using variable feature
Figure 2:c The extracted water segments using enturpy feature
Figure 3: Classification map of land use in flooded area
|
|
|
|
|
|
|
