Supervised Wishart 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_supervised

# 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")

# Hand drawn regions
label_path = Path("../data/sample_labels_sf_alos1.nc")

import os from pathlib import Path import xarray as xr from polsarpro.io import open_netcdf_beam from polsarpro.classification import wishart_supervised # 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") # Hand drawn regions label_path = Path("../data/sample_labels_sf_alos1.nc")

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 [ ]:

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

# initial regions labeled as {1...N}, 0 for unlabelled
labels = xr.open_dataarray(label_path)

# uncomment to test on S matrix made with SNAP S = open_netcdf_beam(input_alos_data) # initial regions labeled as {1...N}, 0 for unlabelled labels = xr.open_dataarray(label_path)

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.

In [3]:

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

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

with ProgressBar():
    wishart_supervised(S, training_labels=labels, boxcar_size=[7, 7]).to_netcdf(file_out)

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

[########################################] | 100% Completed | 11.84 s
[########################################] | 100% Completed | 19.13 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:

In [15]:

from matplotlib.colors import ListedColormap

# Blue: sea, green: vegetation, orange: built-up
colors = ["#387EB9", "#4EAF49", "#FF7F00"]
# Create the discrete colormap
my_cmap = ListedColormap(colors)

res.wishart_supervised_class.plot.imshow(cmap=my_cmap)

from matplotlib.colors import ListedColormap # Blue: sea, green: vegetation, orange: built-up colors = ["#387EB9", "#4EAF49", "#FF7F00"] # Create the discrete colormap my_cmap = ListedColormap(colors) res.wishart_supervised_class.plot.imshow(cmap=my_cmap)

Out[15]:

<matplotlib.image.AxesImage at 0x7d57c4e513d0>