I am processing ALOS PALSAR Full-Polarization data for the application of detection and searching about buried subsurface features or structures near the archaeological sites (Fences, Walls, mounds).
Is this workflow suitable?
1- Calibration radiometrically (Save as Complex Output.
2- Multilooking one time in range and 5 in azimuth.
3- Polarimetric Matrix Generation (T3).
4- Polarimetric Speckle Filtering (Refined Lee Filtering 7*7 Window size).
5- Polarimetric Decomposition with (Freeman Durden)Method.
6- Geometric Terrain Correction with Range Doppler.
7- Showing the map with Scatter Plot.
Calibration Multilook Pol_Martix Pol_SpeckleFiltering PolDecomposition TC.
Any notes are welcomed and highly appreciated.
As the speckle filtering and decomposition are already based on larger kernels, you might skip the multi-looking at the beginning to maintain some fine signals at high resolution.
The largest challenge is probably finding the polarimetric speckle filter which conserves these features best. Do you have field reference data available to see where these structures are located? This would help to evaluate the effect of the filters.
Are you using ALOS or ALOS-2 by the way?
Thanks, Mr. @ABraun
If I understand your post, Firstly my data are ALOSPALSAR-1, downloaded from ESA.
Near the study area I saw surface mounds (Tells in local term) near the archaeological sites. There are some exposed walls as a result of excavation investigations, which are near the surface, about 15 to 10 cm away from the land surface.
Your suggestion is to skip the multilooking process , what about the resolution of backscatter sigma data in the finally, which it become 49 meter while I need it for 10-20 m resolution.?
the first ALOS PALSAR offers only a resolution of 25 - 100 m for quad-pol data (source), so there is nothing at the processing level you can do to reduce it.
Before Terrain Correction, ALOS PALSAR (but not ALOS-2) requires deskewing (example in this tutorial), but I recommend to use SRTM 1Sec (AutoDownload) for this task (and TC), because SRTM 3Sec has problems lately.
So the resolution of ALOSPALSAR when I process ended in 49 m, can I resampled to 25m.
What about the workflow above-mentioned for detection target? Or Can suggest another workflow.
I welcomed your advice.
You can resample it to any resolution, but the information content will remain the same.
If you want to identify subsurface features, you have to find a preprocessing strategy which highlights the ones you already know and then search for similar patterns in the entire image.
If you cannot identify the known ones there are little chances that you will find others, because you don’t know what to look for.