|
|
|
Forest/Resources
|
An Evaluation of Multi-Band/Multipolarised SAR Data For Vegetation Discrimination, In Malaysia
Multifrequencies and polarizations
Table 1 shows the backscattering coefficients ( s0 ) obtained for L and P bands and multi polarization for different test field .
Table 1: Multifrequencies and Multipolarizaiton of the backscattering coefficient
Figure 3: Graph (a) shows the polarization for L band, (b) the polarization of p band and (c) shows the polarization different of L and P bands for natural forest, plantation forest, oil palm and rubber.
The backscatter values were calculated as the mean values taken over all vegetation at the matured stage. Plant moisture content were assumed similar and negligible. The soil moisture is considered as the same for the study area due to flat area due to flat area and very fine weather during the flight flown .
Backscettering value for L and P bands can be seen fro the Tables 1 and Figure 3. for L and, the highest dB values is belonging to suburban area and the lowest values is clear-cut, while the vegetation areas are in the middle of these two. This scenario occurs as well for P band. Base on these phenomena, we could easily discriminate the clear-cut and suburban area, due to the differences off more than 3 dB at any frequencies and polarizations. Diane l Evans et at 1988, also found that the contrast polarization between vegetated area and clear-cut. For vegetation, the dB value is quite close see table 1. le toan et. Al 1997 used 3 dB as a threshold to discriminate rice field. To discriminate oil palm and rubber, L band is the best, where the dB different is more than 3 at any polarizations. The natural forest and plantation forest, is quite difficult to discriminate, because the difference of the dB value is less then 3 at any frequencies and polarizations. However, discrimination can be done to oil palm and plantation forest at L band for polarization VV, HV and TP. Rubber cannot be separated with natural and plantation forest. It might be due to similar characteristics of the vegetation and undergrowth at that area.
Conclusion and Remarks
The Aisar data with multi-frequencies and multi-polarization can discriminate the vegetated an no-vegetated area easily. Discrimination can be done for vegetation with different leaf and trunk structure ,like oil palm and rubber, while rubber, natural forest and plantation forest can not be discriminated. Identified frequencies and polarizations can do classification by segmentation for the vegetation . combination with C band and quantitative validation of vegetation area can further completer the study. Knowledge-based, expert system or neural network should be applicable to refine and enhance this study.
Acknowledgements
The authors hereby acknowledge the contribution given by Mr Nik Nasruddin Mahmood for the completion of this paper.
References :
- A . Rosenquist, 1996: "Evaluation of JERS-1, ERS-1 and Almaz SAR Backscatter for rubber and oil palm stands in West Malaysia. International Journal of Remote Sensing Volume 17, November 16, pp 3219.
- Chen K.S, Huang, W.P., Tsay, D.H. and Amar F, 1996." Classification of Multifrequency Polar metric SAR Imagery Using a Dynamic Learning Neural Network". ". IEEEE Transactions on Geoscience and Remote Sensing, Vol 34, No 3, May 1996, pp 814-820
-
Diane L. Evans, Tom G Farr, Jakob J. Van Zyl and Howard A Zebker, 1988. " Radar Polarimetry : Analysis Tools and Application ". IFEEE Transactions on Geoscience and Remote Sensing, Vol 26, No 6, November 1988, pp 774-789.
-
Huy Le Toan, Florence Ribbes, Li-Fang Wang , Nicolas Floury, Kung-Hau ding, Jin Au Kong, 1997. " Rice Crop Mapping and Monitoring ERS-1 Data Based on Experiment and Modeling Results ". IEEE Transactions on Geosciences and Remote Sensing, Vol 35, No 1, January 1997, pp 41-56.
|
|
|
|
|
|
|
