Wishart H / A / Alpha classification¶

This notebook shows how to apply the Wishart unsupervised classification to a PolSARpro NetCDF dataset.

Import packages and set directories¶

In [1]:

import os 
from pathlib import Path
import xarray as xr
from polsarpro.io import open_netcdf_beam
from polsarpro.classification import wishart_h_a_alpha

# optional import for progress bar
from dask.diagnostics import ProgressBar

# change to your data paths
# original dataset
input_alos_data = Path("/data/psp/test_files/SAN_FRANCISCO_ALOS1_slc.nc")
output_dir = Path("/data/psp/res")

import os from pathlib import Path import xarray as xr from polsarpro.io import open_netcdf_beam from polsarpro.classification import wishart_h_a_alpha # optional import for progress bar from dask.diagnostics import ProgressBar # change to your data paths # original dataset input_alos_data = Path("/data/psp/test_files/SAN_FRANCISCO_ALOS1_slc.nc") output_dir = Path("/data/psp/res")

Load data¶

We load the SNAP NetCDF-BEAM dataset using the open_netcdf_beam function. To obtain such a dataset, please refer to the "Getting Started" tutorial or the quickstart-tutorial.ipynb notebook.

In [2]:

# uncomment to test on S matrix made with SNAP
S = open_netcdf_beam(input_alos_data)

# uncomment to test on S matrix made with SNAP S = open_netcdf_beam(input_alos_data)

Apply the classification¶

Let's apply the algorithm and write the result to a NetCDF file. Optionally we can use a progress bar to monitor the progress of the computation.

Note: Unlike the C implementation, this version does not require precomputed H/A/Alpha parameters. However, for backward compatibility, you may provide existing parameters via the h_a_alpha_result argument. In that case, the H/A/Alpha parameters may have been computed using a different boxcar window size than the one used in this function.

Memory note: On large products, running the classifier without h_a_alpha_result can require substantial memory when the result is computed or written, even with tol_pct=None, because the H/A/Alpha decomposition is part of the same Dask pipeline. If memory is limited, compute the H/A/Alpha decomposition first, save it to NetCDF, reopen it with xarray, and pass it as h_a_alpha_result.

In [3]:

# change to the name of your liking
file_out = output_dir / "psp_wishart_haa.nc"

# netcdf writer cannot overwrite
if os.path.isfile(file_out):
    os.remove(file_out)

with ProgressBar():
    wishart_h_a_alpha(S, boxcar_size=[7, 7]).to_netcdf(file_out)

# change to the name of your liking file_out = output_dir / "psp_wishart_haa.nc" # netcdf writer cannot overwrite if os.path.isfile(file_out): os.remove(file_out) with ProgressBar(): wishart_h_a_alpha(S, boxcar_size=[7, 7]).to_netcdf(file_out)

[########################################] | 100% Completed | 264.04 s

Display outputs¶

We open the previously saved dataset:

In [4]:

res = xr.open_dataset(file_out)

res = xr.open_dataset(file_out)

Then we may plot the individual outputs:

8 Classes based on H/Alpha¶

In [8]:

res.wishart_h_alpha_class.plot.imshow(cmap="Dark2")

res.wishart_h_alpha_class.plot.imshow(cmap="Dark2")

Out[8]:

<matplotlib.image.AxesImage at 0x7691d132d160>

16 Classes based on H/A/Alpha¶

In [ ]:

res.wishart_h_a_alpha_class.plot.imshow(cmap="tab20")

res.wishart_h_a_alpha_class.plot.imshow(cmap="tab20")

<matplotlib.image.AxesImage at 0x7691d112f080>