The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B7. Beijing 2008
308
visual results of weighted RX using simulated image with
various SNR. This algorithm is selected, because it has more
sensitivity of noise increase.
Figure 8. SNR 30:1(a), SNR 20:1(b), SNR 10:1(c), SNR 5:1(d)
Table (2) and figure (9) show the effect of Signal-to-noise ratio
on detection performance (AUC).
Method
SNR = 5
SNR = 10
SNR =20
SNR = 30
Global RX
0.8528
0.9095
0.9285
0.9686
Local RX_UTD
0.8568
0.8806
0.9189
0.9801
Global Weighted
RX-UTD
0.6575
0.7210
0.8326
0.9946
DWEST
0.6711
0.8689
0.9642
0.9719
NSWTD
0.7863
0.8364
0.9061
0.9731
CFT
0.7010
0.8724
0.9210
0.9823
Table 2. AUC in various SNR
Figure 9. Effect of SNR on AUC
Also table (3) and figure (10) show the effect of signal-to-noise
ratio on number of correct detections.
Method
SNR = 5
1
It
SNR = 20
SNR = 30
Global RX
38
44
49
53
Local RXUTD
37
38
41
64
Global Weighted
RX-UTD
3
11
29
61
DWEST
14
34
61
63
NSWTD
24
33
45
64
CFT
17
40
49
60
Table 3. Number of correct detections in various SNR
Effect of Signal To Noise Ratio on Correct Detection
Figure 10. Effect of SNR on correct detection
4. CONCLUSION
Finally our experimental results show that weighted RX-UTD is
the highest performance algorithm (maximum AUC), NSWTD
method has maximum number of correct detections, DWEST-
RX method has minimum number of false alarms, global RX
algorithm has less sensitivity of various SNR and NSWTD is
the highest speed method and it is suitable for processing of
ultraspectral images.
REFERENCES
Chang, C., 2003. Hyperspectral Imaging: Techniques for
Spectral Detection and Classification. Kluwer
Academic/Plenum, New York, pp. 89-02.
Chang, C., 2002. Anomaly detection and classification for
hyperspectral imagery. IEEE, 40(6), pp. 1314-1325.
Hsuan, R., 2005. Weighted anomaly detection for hyperspectral
remotely sensed images. SPIE, 5995(1).
Hsueh, M., 2004. Adaptive causal anomaly detection for
hyperspectral imagery. IEEE. 5(20), pp. 3222-3224.
Kwon, H., 2003. Adaptive anomaly detection using subspace
separation for hyperspectral imagery. Optical Engineering,
42(11), pp. 3342-3351.
Liu, W., 2004. A nested spatial window-based approach to
target detection for hyperspectral Imagery. IEEE, 1(20), pp.
264-268.
Rosario, D., 2005. A nonparametric F-distribution anomaly
detector for hyperspectral imagery. IEEE, 5(12), pp. 2022-2029.