%%capture #Librerías para tratamiento de datos !pip install numpy !pip install pandas #Librería para visualización de imágenes !pip install matplotlib !pip install plotly #Librería para lectura de archivos envi !pip install spectral

#Librerías para tratamiento de datos import numpy as np import pandas as pd #Librería para visualización de imágenes import matplotlib.pyplot as plt from plotly.subplots import make_subplots import plotly.graph_objects as go import plotly.express as px #Librería para lectura de archivos envi import spectral.io.envi as envi from spectral import * #from skimage import io

git_token = "glpat-6mmhJ5Y39dfXnpEwjcwv" !git clone https://JhonnyOsorio:{git_token}@gitlab.com/JhonnyOsorio/Hyperspectral_Images_Analysis

#Reference image suelos path_reference_hdr_suelos = "Hyperspectral_Images_Analysis/suelos/suelo_reference.hdr" path_reference_raw_suelos = "Hyperspectral_Images_Analysis/suelos/suelo_reference.raw" #Sample image suelos path_sample_hdr_suelos= "Hyperspectral_Images_Analysis/suelos/suelo_sample.hdr" path_sample_raw_suelos = "Hyperspectral_Images_Analysis/suelos/suelo_sample.raw" #Sample paleta path_sample_hdr_paleta = "Hyperspectral_Images_Analysis/paleta/paleta.hdr" path_sample_raw_paleta = "Hyperspectral_Images_Analysis/paleta/paleta.raw" #Sample_hojas path_sample_hdr_hojas = "Hyperspectral_Images_Analysis/hojas/hojas.hdr" path_sample_raw_hojas = "Hyperspectral_Images_Analysis/hojas/hojas.raw" #Sample paleta path_sample_hdr_hoja_españa = "Hyperspectral_Images_Analysis/hojas_españa/hoja_españa.hdr" path_sample_raw_hoja_españa = "Hyperspectral_Images_Analysis/hojas_españa/hoja_españa.raw"

def features(path_hdr): info_img= envi.read_envi_header(path_hdr) features = ["Spectral Bands","Resolution","Points"] values = [info_img["bands"],info_img["samples"] + "x" + info_img["lines"],int(info_img["samples"])*int(info_img["bands"])*int(info_img["lines"])] df_feature = pd.DataFrame({"features":features,"values":values}) return df_feature.style.hide_index() def wavelength(path_hdr): #espectro_visible =[] #infrarojo = [] info_img= envi.read_envi_header(path_hdr) wavelength = [str(int(float(i))) for i in info_img["wavelength"]] bands = [i for i in range(len(wavelength))] data = pd.DataFrame({"bands":bands,"wavelength":wavelength}) return data.style.hide_index()

#path_reference_hdr #path_sample_hdr features(path_sample_hdr_hojas)

wavelength(path_sample_hdr_hojas)

def show_static_images(path_hdr,path_raw,snip=(0,None,0,None),cols=5): img = envi.open(path_hdr,path_raw) info_img= envi.read_envi_header(path_hdr) rows = round(int(info_img["bands"])/cols) image = img[snip[0]:snip[1],snip[2]:snip[3],:] wavelength = [str(int(float(i))) for i in info_img["wavelength"]] images = [] for i in wavelength: globals()["img_"+i] = img.read_band(wavelength.index(i))[snip[0]:snip[1],snip[2]:snip[3]] images.append("img_"+i) fig, axs = plt.subplots(rows, cols, figsize=(15, 3*rows)) L_rows = [] for i in range(rows): L_rows += cols*[i] L_cols = rows*list(range(cols)) for i in images : axs[L_rows[images.index(i)],L_cols[images.index(i)]].imshow(eval(i)) axs[L_rows[images.index(i)],L_cols[images.index(i)]].set_title("λ ="+i[4:])

show_static_images(path_reference_hdr_suelos,path_reference_raw_suelos,cols=5)

show_static_images(path_sample_hdr_hojas,path_sample_raw_hojas,cols=5)

def show_interactive_images(path_hdr,path_raw,snip=(0,None,0,None)): img = envi.open(path_hdr,path_raw) info_img= envi.read_envi_header(path_hdr) data = img[:,:,:] fig = px.imshow(data,binary_string=True,animation_frame=2, labels=dict(animation_frame="wavelength"),color_continuous_scale='viridis',range_color=[0,255]) fig.update_traces(hovertemplate="x: %{x} <br> y: %{y} <br> color: %{color[0]}") fig.show()

show_interactive_images(path_reference_hdr_suelos,path_reference_raw_suelos)

show_interactive_images(path_sample_hdr_hojas,path_sample_raw_hojas)

def show_images_wavelength(path_hdr,path_raw,wavelength,snip=(0,None,0,None),interactive =False): img = envi.open(path_hdr,path_raw) info_img= envi.read_envi_header(path_hdr) wl = [str(int(float(i))) for i in info_img["wavelength"]] imagen = img.read_band(wl.index(wavelength))[snip[0]:snip[1],snip[2]:snip[3]] if interactive == False: return plt.imshow(imagen) else: return px.imshow(imagen,color_continuous_scale='viridis')

show_images_wavelength(path_sample_hdr_hojas,path_sample_raw_hojas,wavelength="782",interactive=True)

def unfolding(path_hdr,path_raw,snip=(0,None,0,None), matrix= False): array_data = [] img = envi.open(path_hdr,path_raw) info_img= envi.read_envi_header(path_hdr) wavelength =[str(int(float(i))) for i in info_img["wavelength"]] images = [] for i in wavelength: globals()["img_"+i] = img.read_band(wavelength.index(i))[snip[0]:snip[1],snip[2]:snip[3]] images.append("img_"+i) for i in images: col = eval(i).flatten() array_data.append(col) data = np.stack(array_data, axis=1) df = pd.DataFrame(data).set_axis(wavelength, axis=1) if matrix == True: return data else: return df

unfolding(path_reference_hdr_suelos,path_reference_raw_suelos)

unfolding(path_sample_hdr_hoja_españa,path_sample_raw_hoja_españa)

def wavelength_histogram(path_hdr,path_raw,wavelength): img = envi.open(path_hdr,path_raw) info_img= envi.read_envi_header(path_hdr) wl = [str(int(float(i))) for i in info_img["wavelength"]] banda = wl.index(wavelength) save_rgb(wavelength+".png",img.read_band(banda),format='png') data = io.imread(wavelength+".png") fig = make_subplots(1, 2) # We use go.Image because subplots require traces, whereas px functions return a figure fig.add_trace(go.Image(z=data), 1, 1) data_wavelength = unfolding(path_hdr,path_raw)[wavelength] fig.add_trace(go.Histogram(x=data_wavelength,marker_color="#636EFA"), 1, 2) fig.update_layout(height=450) fig.show()

wavelength_histogram(path_sample_hdr_hojas,path_sample_raw_hojas,wavelength = "782")

import plotly.express as px def histogram_wavelength(path_sample_hdr,path_sample_raw,wavelength): info_img= envi.read_envi_header(path_sample_hdr) wavelength = wavelength data_wavelength = unfolding(path_sample_hdr,path_sample_raw)[wavelength] fig = px.histogram(data_wavelength, x=wavelength) fig.show()

histogram_wavelength(path_sample_hdr_hojas,path_sample_raw_hojas,wavelength = "782")

def clusters(path_hdr,path_raw,snip=(0,None,0,None),k=2,i=3, k_means= False): img = envi.open(path_hdr,path_raw) image = img[snip[0]:snip[1],snip[2]:snip[3],:] info_img= envi.read_envi_header(path_hdr) wavelength = info_img["wavelength"] (m, c) = kmeans(image,k,i) fig = go.Figure() for i in range(c.shape[0]): fig.add_trace(go.Scatter(go.Scatter(x =wavelength, y =c[i] ))) if k_means == False: return px.imshow(m,color_continuous_scale='viridis') else: return fig.show()

clusters(path_sample_hdr_hojas,path_sample_raw_hojas,k=2,i=10)

clusters(path_sample_hdr_hojas,path_sample_raw_hojas,k=2,i=10,k_means= True)

def unfolding_filter(path_sample_hdr,path_sample_raw,wavelength,intensity,black_background=True,matrix=False): data =unfolding(path_sample_hdr,path_sample_raw,matrix = True) info_img= envi.read_envi_header(path_sample_hdr_hojas) wl = [str(int(float(i))) for i in info_img["wavelength"]] banda = wl.index(wavelength) sd = [] if black_background==True: condition = "data[i,banda] < intensity" else: condition = "data[i,banda] > intensity" for i in range(len(data)): if eval(condition): sd.append(i+1) else: sd.append(0) dep_data = [] for band in range(int(info_img["bands"])): sub_data = [] for delt in range(len(sd)-1): if sd[delt] == 0: sub_data.append(data[delt,band]) dep_data.append(sub_data) data_stack = np.stack(dep_data, axis=1) df_depured = pd.DataFrame(data_stack).set_axis(wl, axis=1) Data_Set_Reconstruction = [] for band in range(int(info_img["bands"])): ReSub_Data_Set = [] for delt in range(0, len(sd)): if sd[delt] == 0: ReSub_Data_Set.append(data[delt,band]) else: ReSub_Data_Set.append(0) Data_Set_Reconstruction.append(ReSub_Data_Set) if matrix == True: return np.transpose(Data_Set_Reconstruction) else: unfolding_filter = np.transpose(Data_Set_Reconstruction) df_unfolding_filter = pd.DataFrame(unfolding_filter).set_axis(wl, axis=1) return df_unfolding_filter

unfolding_filter(path_sample_hdr_hojas,path_sample_raw_hojas,wavelength="920",intensity= 60 )

def image_without_background(path_sample_hdr,path_sample_raw,band,wavelength,intensity,black_background=True): Data_Set_Reconstruction = unfolding_filter(path_sample_hdr=path_sample_hdr,path_sample_raw=path_sample_raw,wavelength=wavelength,intensity=intensity,black_background=black_background,matrix=True) info_img= envi.read_envi_header(path_sample_hdr) wl = [str(int(float(i))) for i in info_img["wavelength"]] j = wl.index(band) info_img= envi.read_envi_header(path_sample_hdr_hojas) n = (np.transpose(Data_Set_Reconstruction).shape)[1]/int(info_img["samples"]) imgR = [] for i in range(int(info_img["bands"])): imgr = np.stack(np.array_split(np.transpose(Data_Set_Reconstruction)[i],n)) imgR.append(imgr) return px.imshow(imgR[j],color_continuous_scale='viridis')

image_without_background(path_sample_hdr_hojas,path_sample_raw_hojas,band = "920",wavelength="920",intensity= 60)

def reciprocal(i): if i == 0: return i if i !=0: return 1/i def matrix_normalization(matrix_sam,matrix_ref): reci = np.vectorize(reciprocal) matrix_reciprocal = reci(matrix_ref) matrix_hadamard = np.multiply(matrix_sam,matrix_reciprocal) return matrix_hadamard def dataframe_normalization(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw,snip=(0,None,0,None),matrix=False): info_img= envi.read_envi_header(path_sample_hdr,) wl = [str(int(float(i))) for i in info_img["wavelength"]] matrix_sample = unfolding(path_sample_hdr,path_sample_raw ,matrix = True) matrix_reference = unfolding(path_reference_hdr,path_reference_raw,matrix = True) matrix_normalized = matrix_normalization(matrix_sample,matrix_reference) if matrix == True: return matrix_normalized else: dataframe = pd.DataFrame(matrix_normalized).set_axis( wl , axis=1) return dataframe

dataframe_normalization(path_sample_hdr_suelos,path_sample_raw_suelos,path_reference_hdr_suelos,path_reference_raw_suelos)

def histogram_normalization(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw,wavelength): info_img= envi.read_envi_header(path_sample_hdr) wavelength = wavelength data_wavelength = dataframe_normalization(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw)[wavelength] fig = px.histogram(data_wavelength, x=wavelength) fig.update_xaxes(range=[0,1.5]) fig.show()

histogram_normalization(path_sample_hdr_suelos,path_sample_raw_suelos, path_reference_hdr_suelos,path_reference_raw_suelos,"920")

def normalized_image(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw,band,cube=False): Data_Set_Reconstruction = dataframe_normalization(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw,matrix=True) info_img= envi.read_envi_header(path_sample_hdr_hojas) n = (np.transpose(Data_Set_Reconstruction).shape)[1]/int(info_img["samples"]) imgR = [] for i in range(int(info_img["bands"])): imgr = np.stack(np.array_split(np.transpose(Data_Set_Reconstruction)[i],n)) imgR.append(imgr) data = np.stack(imgR, axis=2) if cube == True: return data else: info_img= envi.read_envi_header(path_sample_hdr) wl = [str(int(float(i))) for i in info_img["wavelength"]] j = wl.index(band) return px.imshow(imgR[j],color_continuous_scale='viridis')

%%javascript IPython.OutputArea.auto_scroll_threshold = 9999 normalized_image(path_sample_hdr_suelos,path_sample_raw_suelos, path_reference_hdr_suelos,path_reference_raw_suelos,band = "920")

def show_interactive_images(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw): info_img= envi.read_envi_header(path_sample_hdr) data = normalized_image(path_sample_hdr,path_sample_raw,path_reference_hdr,path_reference_raw,band = "920",cube=True) fig = px.imshow(data,binary_string=True,animation_frame=2, labels=dict(animation_frame="wavelength"),color_continuous_scale='viridis',range_color=[0,1]) fig.update_traces(hovertemplate="x: %{x} <br> y: %{y} <br> color: %{color}") fig.show()

show_interactive_images(path_sample_hdr_suelos,path_sample_raw_suelos, path_reference_hdr_suelos,path_reference_raw_suelos)