I am working on fusion of SAR and multispectral data for land use land cover classification. I have HH and HV polarized SAR images and multispectral images in the VNIR band. I would like to know the difference in the response of HH and HV polarized SAR data(i.e. how different earth features are discriminated/represented in HH and HV SAR data?). My area of interest consists of natural as well as urban(man-made) features. Which type of SAR data (HH/HV) will be suitable for fusion?
HH measures the proportion of horizontally transmitted waves which return horizontally to the sensor.
HV measures the part of the emmitted waves which are polarized at the earth's surface and return vertically to the sensor. This doesn't happen to a large degree. Accordingly, the measure of HV is way lower than of HH in most cases.
Still, the information content of both rasters ist different. Surfaces which are likely to cause change in polarization are often characterized by volume scattering (branches, dry soil bodies, canopies...).
Comparing HH and HV often shows the spatial distribution of volume scattering in your image. Also, vertically oriented structures (with respect to the looking angle) at the scale of the wavelength of the sensor are highlighted in HV.
There are nice slides provided by the EO College:
https://eo-college.org/resources/polsar/
Very extensive slides provided by ESA:
https://earth.esa.int/documents/10174/669747/PolSARBasic_Concepts.pdf
Hello Sir,
HH and HV polarizations have their own properties according the surface. Urban identification from the SAR data requires HH polarization in the Red Channel. Urban gives the single as well as double (majorly) bounce signature in the SAR signal.
Vegetation identification from the SAR data most preferably uses the HV polarization. We can use the VH polarization as well in the absence of HV.
For your task of fusing SAR and multispectral data for land use land cover classification, you may want to consider both HH and HV polarization data. HH-polarized data are useful for getting information about surface roughness and structure, as like-polarization is sensitive to geometrical characteristics. This could be beneficial for discriminating between different types of man-made structures and possibly types of vegetation as well.
On the other hand, HV-polarized data are useful for volume scattering analysis as it is sensitive to the distribution of scatterers within a medium. This can be beneficial for discriminating different vegetation types and conditions as they have a volumetric structure that causes multiple scattering, leading to a change in polarization.
The fusion of both HH and HV could provide a more comprehensive understanding of both the surface and volume characteristics of the land cover. This, when combined with multispectral data, could potentially lead to a more accurate classification.