Microwave Remote Sensing for Land cover identification
Results and Discusion
It was found that the variation of intensity values of all selected images were reduced on the filtered SAR
Data compare to its original. Figure 3 presents the composite color image of three original SAR data and
figure 4 indicates the same image after filtering with a 5x5window size. The result is that it could reduce
certain noise, cause image is smoother and keep lineament detail on the image. Figure 5 presents the
composite color image after enhancement and some training area allocation. Statistical analysis for
characteristics of the same area for three images taken in almost a dry season showed slight differences
in digital number value..The supervised Maximum Likelihood classified results the six main classes as indicated in Figure 6. The
results of digital analysis confirmed the capability of SAR data used to separate confusion of categories
between village (urban) and water body as well as vegetation canopy condition, however, more details
could be observed and discussed as follows:
Vegetation
Backscatte red signals of SAR data are sensitive to vegetation canopy structure in terms of volume.
Complex canopy reflects high backscattering to the sensor since it gives a strong roughness
characteristic, high signal provides a very bright image. The forest along the river(riparian forest)seems
to be not well separated from the forest. They both appear in bright color.
Compared to its surroundings, it appears in bright tone on SAR image. Paddy and upland crops appear
dark in the
SAR image, but there are very bright parts because there are small crop plants in agricultural field.
However, they are brighter than the water bodies because crop grows large and wetness in the soil
increases radar backscattering.
Water Bodies
Water bodies have a smooth specular surface, hence backscattered signal is very low or no
backscattering to the sensor. Consequently, they appear in dark tone on the SAR image.
Village or Urban
In village/urban, very rough object, backscattering detected is very high since its composition is complex.
Hence urban areas appear in very bright tone .SAR values in urban are a are equally high in all seasons.
Lineament and Other Man Made Feature
The lineament features, such as roads and canals were clearly seen as dark line. Air field, as it is flat
objects, appears in dark zone.
Classification Separability
Resul ts of the overall classification separability could be generally assessed classification in accuracy.
The separability matrix is summarized in Table 2.
Table 2 Separability matrix for land cover classification.
2.00 = very well separated between two classes; 1.90 = well separated between two classes < 1.70 = low separated between two classes
Land Cover Assessment
The overall results of land cover classification using Maximum Likelihood were calculated and
summarized in Table 3.
Table 3 Area and percentage of land cover in the study area
Conclusion
The results of this preliminary study showed that multi -temporal JERS-1 SAR data could be used for
analysis with the expectation of greater advantage for identification of crops types than a single
wavelength image. The general result for land cover classification was fairly obvious, superior at
distinguishing water and village from other categories, but generally suffered for difficulty distinguishing
among forest and riparian forest. Some paddy confused with upland crops. Backscattering is calculated,
village is a highest, second is forest, and the lowest is water bodies. Classification using SAR digital
image found that land cover in this study area mostly contains upland crops and paddy. Agriculture is still
the main important thing for this region. For better results, a deeper understanding of the behavior of the
radar backscatter and its relationship to the biophysical parameter is needed. Note that, bright color
means high backscatter and dark color means low backscatter to the sensor.
In addition, the
electromagnetic interaction with the target is also modified by the object’s dielectric constant. Different
characteristics of radar data, the traditional image analysis developed for conventional optical remotely
sensed data cannot optimally explore the information content of SAR data. A specific algorithm for SAR
classification is needed to be developed..A combination of SAR and optical data would lead to further significant improvement in classification
accuracy and provide additional information, especially the cloud- and haze-affected area. JERS- 1 SAR
is one of the microwave sensors used to monitor land surfaces because of its capability of penetrating
cloud and its high spatial resolution.
Acknowledgement
The author would like to express special thanks to Division of Remote Sensing, National Research
Council of Thailand for providing all the digital JERS-1 SAR data used in this study.
References
- Alves,D.,Skole,D.1996.Characterizing Land Cover Dynamic Using Multi -temporal Imagery, 17, pp. 835 -839.
- Kasisehke,E.S.,Melack ,J.M.,Dobson,M.C.1997. The Use of Imaging Radar for Ecological Application. A
Review Remote Sensing of Environment, 59, pp. 141-156.
- NASDA, 1992. User’s Guide For JERS-1 SAR Data Formal. Tokyo.
- Nasruddlin,N.,et al.1993. SAR and SPOT Data for Land Cover Mapping. Their Complementary and
Integrated Characteristic. Proceeding SAR Data Analysis and Application,Bangkok,Thailand, pp.53- 61.
- Palagonas,M.1993. Land Cover Mapping Using Synthetic Aperture Radar(SAR). Proceedings SAR Data
Analysis and Application, Bangkok Thailand, pp. 175- 192.
