I am working on a research topic, which requires the extraction of RCS values / backscattering coefficients for land surface types. Please let me know if you have any knowledge on how to extract these values from Sentinel 1 SAR data. I want to understand the methodology.
I want to generate a Geospatial layer out of these values for example. in a .tiff file as a attribute table.
Is classification required to get these values for land use types ? Kindly help.
RCS is another name for Sigma0: RCS of Sentinel image
What sort of pre-processing is required before doing the calibration ?
how to extract the Backscatter coefficient values as a table for each classes ? Is it also possible to create a Bitmap with these values ?
Please see here: Radiometric & Geometric Correction Workflow
You can create polygons and extract statistics (e.g. average backscatter per class) with the statistics operator or entire subsets as described here: Export of products from SNAP and here Synergetic use of S1 (SAR) and S2 (optical) data and use of analysis tools
I want to perform a classification procedure to extract the Sigma0 values of the Land surface types (Water, Vegetation, and Urban).
Kindly suggest which classification method is best for retrieving this data. Please explain the process.
I need the average values of the land surface type.
I have done the preprocessing steps for the GRD sentinel 1 product.
I noticed that the water in the edges of the study area was not considered and show as Nan.
Have you seen this? Landcover classification with Sentinel-1 GRD
Thank you for the share. will get back if any questions.
I have worked on RF classification with different classes as described in the manual.
How can I know and extract the Sigma0 values alone from these LabeledClasses? Either from SNAP or QGIS?
If the average values of the labeled classes are embedded in the image? which format is suitable to view the RCS values in a radar application? BMP or .tiff?
Please clarify. Thank you.
You can digitize polygons for certain classes or import them as shapefiles and use the Statistics tool. It gives you the average Sigma0 values below the polygons.
I don’t understand your second question. You either calibrate the pixels to Sigma0 (this is the RCS) or you classify them into land cover.
SNAP is the right tool for both.
Thank you for the response. How to assign a polygon to the image (labeled classes) because the statistics tool works by selecting a certain band and I get the following result without digitizing the polygons on the classes. This is what I see when I extract the statistics for the Labeled classes band.
Kindly explain the process of creating polygons for the classified classes and how to export with the class names (Water, Urban, Vegetation) consisting of Sigma values?
The main purpose is to get the RCS values for the Classes, I thought by classification the average value of the classes can be known.
Classifying pixels based on their average backscatter will probably not give you great results.
This tutorial describes how you load vectors into SNAP so you can use them in the statistics (or histogram) tool: Import of data into SNAP
This tutorial explains how you digitize your own vectors: Synergetic use of S1 (SAR) and S2 (optical) data and use of analysis tools
Thank you for your inputs. I have managed to get some results out. Kindly address my below queries.
- Sigma Values:
Please confirm if the Sigma values here in statistics is the RCS of land surface types.
This was done with manually created training datsets and i generated statistics out of it with the pre-processed Sentinel-1 GRD product (TC Stack image - VH).
- RF Classification:
I have done the same process with RF classification image, as u mentioned the results were not great (used same training datasets). I did not include the Image analysis band in the RF classification along with the PCA and TC stack. As, the process was running for so long, i had to cancel it. The resulting output had all the bands but with failed message. Please check the screenshots.
Actually, Sigma is the standard deviation of all values inside the polygon. I think you are looking for the mean.
The process of classification is correct, maybe you can start with less bands and increase the number of input bands step by step. Or you create a spatial subset to reduce the data volume.
- Mean values:
Okay, the mean values look like this.
Why are the backscatter values for all the classes negative? Water should have the least and other in positive right? any idea what is the issue here?
- Image feature analysis
I have tried spatial subset and also for individual bands, end up getting these error messages. Even if it completes, the image is not opening.
The GLCM textures work by changing the Quantizer to Equal Distance Quantizer, instead of a probabilistic quantizer. Can I use these bands for RF classification?
This is completely normal for calibrated radar data in dB. Among all classes, water has the lowest backscatter, so it is has the lowest negative value possible. Please check here: Backscattering coefficient and radar reflectivity
This will be fixed with the next update, please see here: GLCM Error
But basically, any texture is potentially helpful to increase the feature space.
Thank you for the prompt response and information I will try out the RF classification now.