def generateTrainingData(image, surface_class, bandnames):
im = image[2:5, :, :].transpose()
ims = np.zeros([image.shape[1], image.shape[2], 3])
for i, band in enumerate([2, 1, 0]):
ims[:, :, i] = plot.adjust_band(image[band, :, :], kind='linear')
# Get Water data
fig = plt.figure()
t = plt.gca()
im = plt.imshow(ims)
PD = pickData(t)
axclear = plt.axes([0.0, 0.0, 0.1, 0.1])
bclear = Button(plt.gca(), 'Clear')
bclear.on_clicked(PD.clear)
axdone = plt.axes([0.2, 0.0, 0.1, 0.1])
bdone = Button(plt.gca(), 'Done')
bdone.on_clicked(PD.done)
fig.canvas.mpl_connect('button_press_event', PD)
im.set_picker(5) # Tolerance in points
plt.show()
# Convert rectangles to DF
df = pandas.DataFrame()
for rect in PD.rects:
# Get indexes at bottom left
botleft = rect.get_xy()
botleft = [math.ceil(i) for i in botleft]
# Get indexes at top right
topright = [
botleft[0] + rect.get_width(),
botleft[1] + rect.get_height(),
]
topright = [math.ceil(i) for i in topright]
# Get image rows
ys = [botleft[1], topright[1]]
xs = [botleft[0], topright[0]]
sample = image[:, min(ys):max(ys), min(xs):max(xs)]
print(sample.shape)
df = pandas.concat([df, dataBox2df(sample,
bandnames)]).reset_index(drop=True)
df['class'] = surface_class
return df
img[b, :, :][img[4, :, :] == 255] = 255
# Remove map band
img = img[:4, :, :]
#scaled_img = resize(img, (int(scale * float(rows)), int(scale * float(cols))))
# reshape into long 2d array (nrow * ncol, nband) for classification
reshaped_img = reshape_as_image(img)
# Kmeans
k = 8
kmeans_predictions = KMeans(n_clusters=k, random_state=0,
algorithm='full').fit(reshaped_img.reshape(-1, 4))
kmeans_predictions_2d = kmeans_predictions.labels_.reshape(rows, cols)
# Visualize
rgb = img[0:3]
rgb_norm = adjust_band(rgb) # normalize bands to range between 1.0 to 0.0
rgb_reshaped = reshape_as_image(rgb_norm) # reshape to [rows, cols, bands]
fig, axs = plt.subplots(2, 1, figsize=(10, 20))
show(rgb_norm, ax=axs[1])
axs[1].set_title("RGB")
axs[0].imshow(kmeans_predictions_2d)
axs[0].set_title("KMEANS")
plt.show()
sentinel_rgb.bounds
sentinel_rgb.crs
sentinel_rgb.transform
sentinel_rgb.indexes
sentinel_rgb.read(3)
img = sentinel_rgb.read(3)
show(img)
#producing RGB map with sentinel image
from rasterio.plot import show, adjust_band
imgdata = np.array([adjust_band(sentinel_rgb.read(i)) for i in (4,3,2)])
show(imgdata*10) # factor 10 to increase brightness
#NDWI Bands green and NIR
band3 = rasterio.open('Desktop/Dersler/rasterio deneme/sentinel_band3.tif') #green
band8 = rasterio.open('Desktop/Dersler/rasterio deneme/sentinel_band8.tif') #NIR
band3.height
band3.width
band3.crs
band3.transform
band8.height
band8.width
band8.crs
def test_plot_normalize():
a = np.linspace(1, 6, 10)
b = adjust_band(a, 'linear')
np.testing.assert_array_almost_equal(np.linspace(0, 1, 10), b)
def test_plot_normalize():
a = np.linspace(1, 6, 10)
b = adjust_band(a, 'linear')
np.testing.assert_array_almost_equal(np.linspace(0, 1, 10), b)