I am doing a project with Sentinel 1 and I have the two images: SLC and GRD (They are already downloaded, single and dual pole), the purpose of my project is to create a classification for a region where I should detect some coverage as: crops, forest, cities (or urban zones) and wather (river, lakes) but I focus in the SLC images and I have advanced with these types images and I follow the next workflow for the pre-processing SLC images:
Orbite calibration
Calibrate
Deburst
Multilooking
Terrain Correction
Do You think that these steps become a SLC to a GRD image for I can work with the backscattering but since a SLC image ???
Do You explain how detect some coverage that I said before, some steps???
Besides What are the differences between SLC adn GRD images??? and if I use the GRD images What are the steps for processing correctly ??? or Should I do some PRE-processing with the GRD image ??? or this is ready for work in contradistinction to SLC that i have to do a PRE-processing
ABraun
April 21, 2017, 5:49pm
2
About the differences between SLC and GRD, please see here:
it means that its already ground-range corrected. Unlike SLC data the pixels’ areas are already of same size.
Astrium has made nice graphics for this case. They refer to TerraSAR-X but the principle is the same for most SAR satellites:
Slant Range Complex (SLC/SSC/L.1.1)
[image]
Data is neither multi-looked nor projected
Ground Range Detected (GRD/MGD/L1.5)
[image]
Data is multi-looked but not projected
Geocoded Ellipsoid Corrected (GEC/L2.0)
[image]
Data is projected but topographic d…
About suitable workflows:
for SLC data I would suggest
TOPS Split
Apply Orbit file
Thermal Noise Removal
Calibration to Beta0*
TOPSAR Deburst
Radiometric terrain flattening
(Speckle filtering)
Range Doppler Terrain Correction
If you don’t need the phase information you can also download it as a GRD product and only apply the following:
Apply Orbit file
Thermal noise removal
Calibration to Beta0
Radiometric terrain flattening
(Speckle filtering)
Range Doppler Terrain Correction
or (if you don’t have much topography)…
But I think GRD images are enough for classification purposes:
for the estimation of forest areas GRD data is completely sufficient.
The general work flow is:
Apply Orbit File
Calibrate
Speckle Filter
Range-Doppler Terrain Correction
You can use the ratio between VV and VH in order to extract the forest areas.
VH values are high where ever the incoming vertical wave is backscattered horizontally to a high degree. This is often the case for vegetation with larger volumes.
Make A RGB composite out of Red: VV Green: VH and Blue: VH/VV. The Forest ares sh…
How classifications are performed in SNAP is explained here:
I’m looking to do a landcover classification of Sentinel 1 sar data in either SNAP or QGIS. Does anyone know how to go about doing this?
2 Likes
Thanks ABraun, it really is a excellent explication … Maybe Can you the answer of some other questions that i put before ???
For example: Do You explain how detect some coverage that I said before, some steps???
Thanks ABraun, it really is a excellent explication … Maybe Can you the answer of some other questions that i put before ???
For example: Do You explain how detect some coverage that I said before, some steps???
Try with GRD and see if the results are ok. Read the tutorials plus some scientific papers on classification of SAR images.
ABraun
April 21, 2017, 6:04pm
6
if you mean classification: It is explained here (for Sentinel-2, but works as well with Sentinel-1, as long as you have some bands)
if the unsupervised classification took too long you can
subset your image and test it (a whole S2 scene could take really long for a cluster or EM analysis)
reduce the number of clusters
Choose another classifier (EM and K-means take different time)
For a supervised classification you have to define geometries including your training areas. [image]
Make one per class you want to detect.
[image]
They are stored in the vector Data folder (don’t mind about pins and gcps)
[image]
You can t…
1 Like