Sentinel 1 land cover classification

s1tbx
#1

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?

Sentinel 1 images SLC and GRD
#2

Please take a look at this

#3

Thank you so much! I’m afraid I’m not very good at programming. I was hoping there is a tool I could run my image through to generate a classification. I have tried the Texture Analysis Using the Gray-Level Co-Occurrence Matrix (GLCM) in SNAP and have tried the Semi auto automatic classification plug-in in QGIS but my computer keeps crashing, so I’m unsure if these methods work or not. Do you know should these methods work?

#4

If you read the abstract of the article I sent it to you within my previous post carefully, you could extract the procedure or method according to what you want to . implement, and then according to your method you could find out or answer this question yourself, Could I use snap to reach my goal?

#5

Thank you :slight_smile:

#6

this is a short description on image classification in SNAP

If you don’t want to use SNAP you can try the EnMAP toolbox
http://www.enmap.org/?q=enmapbox

1 Like
#8

the methods (collection of samples and training/application of classifiers) work with SAR data as well. Especially the random forest classifier takes input of different units so SAR intensities and textures can be applied.

I would however suggest to convert to Sigma0 db so the data range is roughly between -40 and +5.

1 Like
#9

Hi there

just making sure that I’ve understood this procedure correctly.

First I go wit h this procedure http://sentinel1.s3.amazonaws.com/docs/S1TBX%20SAR%20Basics%20Tutorial.pdf
Second, I convert into Sigma0 db --> How, what steps in SNAP?
Third I go with your proposed "Supervised and unsupervised classification, Sentinel 2

Done! correct?

Kind regards

1 Like
#10

yes. Multi-looking is not necessarily required but improves your data regarding the presence of speckle.

You can convert your data by right-clicking on the bands > “Convert to db”
This generates a virtual raster. Make it physical by right-clicking “convert band” and save your product.

1 Like
#11

Thank you ABraun for that fast answer.
I’m comming from the interferometry field…and was kind of lost in this topic.
Thank you and have a nice weekend.

#12

Good luck. You can share your results in here with screenshots and we can maybe help again if you are stuck somewhere.

#13

I am having my training data sets as point file. Is it possible to classify S1 data in SNAP using points?

#14

I don’t think so. In any case for SAR data using a single pixel would be bad.

1 Like
#15

If I have the polygons, is it possible to do supervised classification in S1 images with SNAP?

Organic crops with Sentinel 1 data
#16

yes. Open your SAR image, menuy > import > other > shapefile
Then use the polygons as training vectors in the classification modules by selecting the column that has the class information.

1 Like
Classification of Sentinel 1A imagery
#17

How do you execute a Support Vector Machines classification to map rice areas in Sentinel-1 toolbox? thanks before

#18

SVM is currently not implemented in SNAP

1 Like
#19

Thanks for the reply ABraun. Do you have any suggestions on what classification method i can use to map rice areas that can be achieved with SNAP Sentinel-1 toolbox? My data is IW GRDH by the way

#20

the random forest classifier is quite good if you have data of different scales. But it needs more than just 2 rasters.
For the case you only have VV and VH you can use the KD Tree KNN classifier.

1 Like
#21

Okay Abraun how do you execute a KD Tree KNN classifier on SNAP? I’ve tried looking at http://step.esa.int/main/doc/tutorials/ and the ESA training courses but couldn’t find what was needed. Sorry for the many questions.