Add missed 'utils' file

pygmtsar/pygmtsar/utils.py

@@ -0,0 +1,107 @@
+# ----------------------------------------------------------------------------
+# PyGMTSAR
+# 
+# This file is part of the PyGMTSAR project: https://github.com/mobigroup/gmtsar
+# 
+# Copyright (c) 2023, Alexey Pechnikov
+# 
+# Licensed under the BSD 3-Clause License (see LICENSE for details)
+# ----------------------------------------------------------------------------
+class utils():
+
+#     @staticmethod
+#     def regression(phase, topo, fit_intercept=True):
+#         import numpy as np
+#         import xarray as xr
+#         from sklearn.linear_model import LinearRegression
+#         from sklearn.pipeline import make_pipeline
+#         from sklearn.preprocessing import StandardScaler
+# 
+#         # define on the same grid
+#         topo = topo.reindex_like(phase, method='nearest')
+#     
+#         # build prediction model with or without plane removal (fit_intercept)
+#         regr = make_pipeline(StandardScaler(), LinearRegression(fit_intercept=fit_intercept))
+# 
+#         topo_values = topo.values.ravel()
+#         phase_values = phase.values.ravel()
+#         nanmask = np.isnan(a) | np.isnan(b)
+# 
+#         # fit on non-NaN values only and predict on the full grid
+#         phase_topo = regr.fit(np.column_stack([topo_values[~nanmask]]),
+#                               np.column_stack([phase_values[~nanmask]]))\
+#                     .predict(np.column_stack([topo_values])).reshape(phase.shape)
+#         return xr.DataArray(phase_topo, coords=phase.coords)
+
+    @staticmethod
+    def histogram(data, bins, range):
+        """
+        hist, bins = utils.histogram(corr60m.mean('pair'), 10, (0,1))
+        print ('bins', bins)
+        hist = dask.compute(hist)[0]
+        print (hist)
+        """
+        import dask
+        return dask.array.histogram(data, bins=bins, range=range)
+
+    @staticmethod
+    def corrcoef(data, mask_diagonal=False):
+        """
+        Calculate the correlation coefficients matrix for a stack of interferograms.
+
+        Parameters:
+        - data (xarray.DataArray): A 3D input grid representing a stack of interferograms.
+                                  The dimensions should include time, 'y', and 'x'.
+        - mask_diagonal (bool): If True, the diagonal elements of the correlation matrix
+                                will be set to NaN to exclude self-correlation.
+
+        Returns:
+        xarray.DataArray: The cross-correlation matrix.
+        
+        Example:
+        # plot variance-covariance matrix
+        corr_stack = corr60m.mean('pair')
+        corr = utils.corrcoef(unwrap.phase.where(corr_stack.where(corr_stack>0.7)))
+        corr.where((corr)>0.7).plot.imshow(vmin=-1, vmax=1)
+        plt.xticks(rotation=90)
+        plt.show()
+        
+        # find the most correlated interferograms to exclude
+        corr = utils.corrcoef(unwrap.phase.where(corr_stack.where(corr_stack>0.7)), mask_diagonal=True)
+        df = corr.where(corr>0.7).to_dataframe('').reset_index().dropna()
+        exclude = np.unique(np.concatenate([df.ref, df.rep]))
+        exclude
+        """
+        import xarray as xr
+        import dask
+        import numpy as np
+        assert len(data.dims) == 3, 'ERROR: expected 3D input grid'
+        stackdim = data.dims[0]
+        stackvals = data[stackdim].values
+        corr = dask.array.corrcoef(data.stack(points=['y', 'x']).dropna(dim='points').data).round(2)
+        corr = xr.DataArray((corr), coords={'ref': stackvals, 'rep': stackvals}).rename('corr')
+        corr['ref'] = corr['ref'].astype(str)
+        corr['rep'] = corr['rep'].astype(str)
+        if mask_diagonal:
+            # set diagonal elements to NaN
+            diagonal_mask = np.eye(len(stackvals), dtype=bool)
+            return corr.where(~diagonal_mask, np.nan)
+        return corr
+
+    @staticmethod
+    def binary_erosion(data, *args, **kwargs):
+        """
+        https://docs.scipy.org/doc/scipy/reference/generated/scipy.ndimage.binary_erosion.html
+        """
+        import xarray as xr
+        from scipy.ndimage import binary_erosion
+        # Perform binary erosion on the NumPy array data
+        array = binary_erosion(data.values, *args, **kwargs)
+        # Create a new DataArray from the eroded NumPy array
+        array = xr.DataArray(
+            array,
+            coords=data.coords,
+            dims=data.dims,
+            attrs=data.attrs,
+        )
+        return array