I have GPS point of various features and i want to use these points as a sample. Please guide me how to use GPS point as a sample in RF classification using SNAP.
If you have them as a Shapefile or CSV you can load them by selecting the product and then File > Import > Vector
Your file however needs a column with the class number or description.
Thanks sir, it worked. But i got some error in classified image i.e. mining and bare soil is mixed with sattlement area so i am nit able to discriminate these features
in this case you either need more input layers or better training samples. Please have a look here:
Sir one more question which polarization should be use VV or VH or VV+VH. Please help.
all of them. The more raster layers you have for training and predicting the higher are the chances for getting better results.
1 Like
Thanks a lot now it worked properly…thank you so much…one more question now i want to do accuracy assessment please help.
Accuracy assessment is not supported in SNAP by now.
Please use the search function for such questions:
Maybe the path is too long, please try C:\temp\test.dim
sir, I am doing the paddy field identification using sentinal 1, a time series analysis
first i have done the preprocessing of sentinal 1 images
second i convert into sigma0 dB
then in time series graph with vv/vh polarisation,i got a backscatter profile of paddy
how to do a random forest classification with the unique profile of paddy in the graph?
the time-series tool shows the temporal domain of a pixel whiel Random Forest is a spatial predictor in SNAP. So you can use several dates as input bands for the classification.
Please have a look at this example: Random Forest Classifier struggles with long names
sir, How to extract the mean DN value from time series SAR images for each polygon.
?
A tutorial on time-series analysis will be published within the next few days.
here it is: Time-series analysis with Sentinel-1