Spectral Library and Cloud Helpers¶

This notebook demonstrates spectral library matching and chunk planning for cloud-native hyperspectral workflows.

In [ ]:

# %pip install hypercoast

# %pip install hypercoast

In [ ]:

from pathlib import Path
from tempfile import TemporaryDirectory

import numpy as np
import xarray as xr

import hypercoast

from pathlib import Path from tempfile import TemporaryDirectory import numpy as np import xarray as xr import hypercoast

Build and save a small spectral library¶

In [ ]:

wavelengths = np.array([450.0, 550.0, 650.0, 750.0, 850.0])
library = hypercoast.SpectralLibrary(
    wavelengths=wavelengths,
    spectra={
        "clear_water": np.array([0.015, 0.025, 0.010, 0.006, 0.004]),
        "turbid_water": np.array([0.030, 0.055, 0.045, 0.030, 0.020]),
        "vegetation": np.array([0.030, 0.060, 0.040, 0.350, 0.500]),
    },
)
library.to_dataframe()

wavelengths = np.array([450.0, 550.0, 650.0, 750.0, 850.0]) library = hypercoast.SpectralLibrary( wavelengths=wavelengths, spectra={ "clear_water": np.array([0.015, 0.025, 0.010, 0.006, 0.004]), "turbid_water": np.array([0.030, 0.055, 0.045, 0.030, 0.020]), "vegetation": np.array([0.030, 0.060, 0.040, 0.350, 0.500]), }, ) library.to_dataframe()

In [ ]:

with TemporaryDirectory() as tmpdir:
    csv_path = Path(tmpdir) / "spectral-library.csv"
    hypercoast.write_spectral_library(library, csv_path)
    reloaded = hypercoast.read_spectral_library(csv_path)
    display(reloaded.to_dataframe())

with TemporaryDirectory() as tmpdir: csv_path = Path(tmpdir) / "spectral-library.csv" hypercoast.write_spectral_library(library, csv_path) reloaded = hypercoast.read_spectral_library(csv_path) display(reloaded.to_dataframe())

Match spectra¶

In [ ]:

observed = np.array(
    [
        [0.016, 0.024, 0.011, 0.006, 0.004],
        [0.029, 0.054, 0.046, 0.031, 0.019],
    ]
)

hypercoast.match_spectra(observed, library, method="sam", top_k=2)

observed = np.array( [ [0.016, 0.024, 0.011, 0.006, 0.004], [0.029, 0.054, 0.046, 0.031, 0.019], ] ) hypercoast.match_spectra(observed, library, method="sam", top_k=2)

In [ ]:

spectrum = observed[0]
continuum_removed = hypercoast.continuum_remove(spectrum, wavelengths)
derivative = hypercoast.derivative_spectrum(spectrum, wavelengths)

continuum_removed, derivative

spectrum = observed[0] continuum_removed = hypercoast.continuum_remove(spectrum, wavelengths) derivative = hypercoast.derivative_spectrum(spectrum, wavelengths) continuum_removed, derivative

Plan cloud-native chunks¶

Use suggest_chunks before opening or exporting large hyperspectral scenes with Dask-backed xarray workflows.

In [ ]:

cube = xr.DataArray(
    np.zeros((128, 120, 160), dtype="float32"),
    dims=("wavelength", "y", "x"),
    coords={"wavelength": np.linspace(400, 900, 128)},
    name="reflectance",
)

hypercoast.suggest_chunks(cube, target_pixels=25_000)

cube = xr.DataArray( np.zeros((128, 120, 160), dtype="float32"), dims=("wavelength", "y", "x"), coords={"wavelength": np.linspace(400, 900, 128)}, name="reflectance", ) hypercoast.suggest_chunks(cube, target_pixels=25_000)