example_analysis_hypercube.py

import os
import sys
from xml.dom.domreg import well_known_implementations

sys.path.append(f"..{os.sep}..")
from core.Analysis import Analysis
import plugins.imaging_methods.FIS_common_functions.FIS_common_analysis as ts

import matplotlib.pyplot as plt
from sklearn.decomposition import PCA
from sklearn.cluster import KMeans
import spectral as sp
import numpy as np


an = Analysis()  # creation of the anaysis class
an.load_data()  # loading of the hypercube by a tkinter windows

fis = (
    ts.FisAnalysis()
)  # creating the specialize FIS analysis class for analysis of datacube
rgb_image = fis.RGB_reconstruction(
    an.reconstructed_data, an.wavelengths
)  #  create a false rendering in RGB domain of the hyeprcube
datacube = fis.datacube_normalisation_snv(an.reconstructed_data)
image_seg = fis.clustering(
    datacube, 3, 3
)  #  segmentation process using kmeans  with 3 clusters on the 3 first dimensions of PCA


fig = plt.figure()
plt.suptitle("Hyperspectral segmentation result")
plt.subplot(1, 4, 1)
plt.imshow(rgb_image)
plt.axis("off")
plt.title("False RGB image of the datacube")
plt.subplot(1, 4, 2)
plt.imshow(image_seg)
plt.title("Segmented datacube")
plt.axis("off")


"""
plotting the result of the segmentation in the PCA domain 
"""
image_reshape = np.reshape(
    datacube, (np.size(datacube, 0) * np.size(datacube, 1), np.size(datacube, 2))
)
pca = PCA(3)
principalComponents = pca.fit_transform(image_reshape)
kmeans = KMeans(3, n_init=11, random_state=0).fit(principalComponents)
kmeans.labels_


ax = fig.add_subplot(1, 4, 3, projection="3d")

ax.scatter(
    principalComponents[np.where(kmeans.labels_ == 0), 0],
    principalComponents[np.where(kmeans.labels_ == 0), 1],
    principalComponents[np.where(kmeans.labels_ == 0), 2],
    "r",
)
ax.scatter(
    principalComponents[np.where(kmeans.labels_ == 1), 0],
    principalComponents[np.where(kmeans.labels_ == 1), 1],
    principalComponents[np.where(kmeans.labels_ == 1), 2],
    "r",
)
ax.scatter(
    principalComponents[np.where(kmeans.labels_ == 2), 0],
    principalComponents[np.where(kmeans.labels_ == 2), 1],
    principalComponents[np.where(kmeans.labels_ == 2), 2],
    "r",
)

plt.title("Datacube pixels in the PCA domain")
plt.xlabel("First PC")
plt.ylabel("Second PC")
ax.set_zlabel("Third PC")


plt.subplot(1, 4, 4)
mean_spectra = []
for cluster in range(np.max(image_seg) + 1):
    mask = image_seg == cluster
    mask = np.tile(mask[:, :, np.newaxis], np.size(datacube, 2))
    mean_spectra.append(np.mean(mask * an.reconstructed_data, axis=(0, 1)))

mean_spectra = np.asarray(mean_spectra)
mean_spectra -= np.mean(mean_spectra[:, :50], 1)[:, np.newaxis]
ref_cluster = 0
normalised_spectra = mean_spectra / mean_spectra[ref_cluster]

plt.plot(an.wavelengths, 0.2 * normalised_spectra[1:, :].T)
plt.ylim(0, 0.5)
plt.xlim(400, 800)


plt.show()