The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
Figure 7. Spectral curve of unidentified substance having the 
similar spectral with vegetable in near-infrared and 
red light waveband 
Interestingly, the value of NDVI and RENDVI of healthy 
vegetable is different from the value of contaminative vegetable. 
Therefore, the contaminated vegetable could be detected by the 
D-NDVI. One major issue for D-NDVI is that there are some 
substances showing in the result of D-NDVI do not belong to 
vegetation labeled “A” and “B” (Figure 9). This is due to the 
fact that the spectral curves of unidentified substance have the 
similar spectral with vegetable in near-infrared and red light 
waveband (Figure 7). In order to address this issue, this 
experiment combines the result of ICA-based vegetable 
information extraction that present the vegetable distribution 
(Figure 5) and the result of RENDVI-NDVI (Figure 8) that 
presents the contaminative information distribution to exclude 
the interference of non-vegetation. 
/ ' 
& 
, * 
A' 
Figure 8. D-NDVI Where A and B present the unidentified 
substance having similar spectral curve with vegetable 
Figure 9. The result of contaminative information distribution 
5. CONCLUSIONS 
Vegetable spectral characteristics-based soil contaminative 
information extraction has found important application in the 
recent years. Several remarkable works has also been reported 
(Jago et al., 1995; Buschmann et al., 1998). But those research 
mainly base on the support of field data to detect contaminative 
information. They have the disadvantage of greater complexity 
and higher cost. Actually, we need to estimate contaminative 
information without field data. Interestingly, if vegetable was 
contaminated, the spectral curve will transform, the red edge 
will shift to blue light waveband. By comparing the spectral 
difference of healthy vegetable and contaminated vegetable and 
analyzing the vegetable contaminative mechanism, this paper 
proposed a new approach that introduce contaminative factor to 
estimate contamination distribution. Nevertheless, there are 
some substance having similar spectral characteristics in object 
waveband (red light and near-infrared waveband) interfere 
contaminative information extraction. To exclude that 
interference, this paper process the image data using Fast-ICA, 
then extract vegetable distribution information based on PPI 
and SAM technology. Finally, extract contaminative 
information by intersecting the result of D-NDVI with the result 
of ICA based-vegetable information extraction. A real Hyperion 
image experiment is conducted to validate the utility in real 
applications. However, one limitation of the results developed 
in this paper is the choice of threshold in step3. For the scarcity 
of field data, it is difficult to calculate the threshold exactly. In 
this experiment threshold is estimated by the spectral difference. 
Further research will be devoted to removing this limitation. 
ACKNOWLEDGMENT 
The authors would like to thank anonymous reviewers for their 
thoughtful comments. This work is under the auspice of 
National High-tech R&D Program of China (863 program) 
(2007AA12Z174) and National Natural Science Foundation of 
China (40771155). 
REFERENCES 
Boochs, F., Kupfer, G., 1990. Shape of the red edge as vitality 
indicator for plants [J], Int J Remote Sensing, 11(10): 1741-1753. 
Buschmann, C., Lichtenthaler, H.K., 1998. Principles and 
characterization of multi-colour fluorescence imaging of plants 
[J], Journal of Plant Physiology, 152:297-314. 
Du, P, Zhao, HJ, Zhang, B, Zheng, LF, 2005. Independent 
component analysis for hyperspectral imagery plant 
classification. Applications of neural networks and machine 
learning in image processing, ix 5673: 71-81. 
Gong, P, Pu, RL, Miller, J.R., 1995. Coniferous forest leaf-area 
index estimation along the Oregon transects using compact 
airborne spectrographic imager data. Photogramm Eng Remote 
Sens 61 (9):1107-1117. 
Hoque, E., Huntzler, J.S., 1992. Spectral blue shift of red edge 
monitors damage class of beech trees [J], Remote sensing of 
environment, (39):81~84. 
Horler, D. H. N, Barber J., 1983. The Red Edge of Plant Leaf 
Reflectance [M]. Int J Remote Sens, 4: 273-288.