forked from kilianv/CoastSat_WRL
testing the NN classifier
kvos
7 years ago
parent
0a78e3c539
commit
220e5557f7
@ -0,0 +1,372 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
|
||||
#==========================================================#
|
||||
# Extract shorelines from Landsat images
|
||||
#==========================================================#
|
||||
|
||||
# Initial settings
|
||||
import os
|
||||
import numpy as np
|
||||
import matplotlib.pyplot as plt
|
||||
import ee
|
||||
import pdb
|
||||
|
||||
# other modules
|
||||
from osgeo import gdal, ogr, osr
|
||||
import pickle
|
||||
import matplotlib.cm as cm
|
||||
from pylab import ginput
|
||||
|
||||
# image processing modules
|
||||
import skimage.filters as filters
|
||||
import skimage.exposure as exposure
|
||||
import skimage.transform as transform
|
||||
import sklearn.decomposition as decomposition
|
||||
import skimage.measure as measure
|
||||
import skimage.morphology as morphology
|
||||
|
||||
# machine learning modules
|
||||
from sklearn.model_selection import train_test_split
|
||||
from sklearn.neural_network import MLPClassifier
|
||||
from sklearn.preprocessing import StandardScaler, Normalizer
|
||||
from sklearn.externals import joblib
|
||||
|
||||
# import own modules
|
||||
import functions.utils as utils
|
||||
import functions.sds as sds
|
||||
|
||||
# some settings
|
||||
np.seterr(all='ignore') # raise/ignore divisions by 0 and nans
|
||||
plt.rcParams['axes.grid'] = True
|
||||
plt.rcParams['figure.max_open_warning'] = 100
|
||||
ee.Initialize()
|
||||
|
||||
# parameters
|
||||
cloud_thresh = 0.5 # threshold for cloud cover
|
||||
plot_bool = False # if you want the plots
|
||||
min_contour_points = 100# minimum number of points contained in each water line
|
||||
output_epsg = 28356 # GDA94 / MGA Zone 56
|
||||
buffer_size = 10 # radius (in pixels) of disk for buffer (pixel classification)
|
||||
min_beach_size = 50 # number of pixels in a beach (pixel classification)
|
||||
|
||||
# load metadata (timestamps and epsg code) for the collection
|
||||
satname = 'L8'
|
||||
sitename = 'NARRA'
|
||||
#sitename = 'OLDBAR'
|
||||
|
||||
# Load metadata
|
||||
filepath = os.path.join(os.getcwd(), 'data', satname, sitename)
|
||||
with open(os.path.join(filepath, sitename + '_timestamps' + '.pkl'), 'rb') as f:
|
||||
timestamps = pickle.load(f)
|
||||
with open(os.path.join(filepath, sitename + '_accuracy_georef' + '.pkl'), 'rb') as f:
|
||||
acc_georef = pickle.load(f)
|
||||
with open(os.path.join(filepath, sitename + '_epsgcode' + '.pkl'), 'rb') as f:
|
||||
input_epsg = pickle.load(f)
|
||||
with open(os.path.join(filepath, sitename + '_refpoints' + '.pkl'), 'rb') as f:
|
||||
refpoints = pickle.load(f)
|
||||
# sort timestamps and georef accuracy (dowloaded images are sorted by date in directory)
|
||||
timestamps_sorted = sorted(timestamps)
|
||||
idx_sorted = sorted(range(len(timestamps)), key=timestamps.__getitem__)
|
||||
acc_georef_sorted = [acc_georef[j] for j in idx_sorted]
|
||||
|
||||
# path to images
|
||||
file_path_pan = os.path.join(os.getcwd(), 'data', satname, sitename, 'pan')
|
||||
file_path_ms = os.path.join(os.getcwd(), 'data', satname, sitename, 'ms')
|
||||
file_names_pan = os.listdir(file_path_pan)
|
||||
file_names_ms = os.listdir(file_path_ms)
|
||||
N = len(file_names_pan)
|
||||
|
||||
# initialise some variables
|
||||
cloud_cover_ts = []
|
||||
date_acquired_ts = []
|
||||
acc_georef_ts = []
|
||||
idx_skipped = []
|
||||
idx_nocloud = []
|
||||
t = []
|
||||
shorelines = []
|
||||
|
||||
#%%
|
||||
for i in range(1):
|
||||
i = 0
|
||||
# read pan image
|
||||
fn_pan = os.path.join(file_path_pan, file_names_pan[i])
|
||||
data = gdal.Open(fn_pan, gdal.GA_ReadOnly)
|
||||
georef = np.array(data.GetGeoTransform())
|
||||
bands = [data.GetRasterBand(i + 1).ReadAsArray() for i in range(data.RasterCount)]
|
||||
im_pan = np.stack(bands, 2)[:,:,0]
|
||||
nrows = im_pan.shape[0]
|
||||
ncols = im_pan.shape[1]
|
||||
# read ms image
|
||||
fn_ms = os.path.join(file_path_ms, file_names_ms[i])
|
||||
data = gdal.Open(fn_ms, gdal.GA_ReadOnly)
|
||||
bands = [data.GetRasterBand(i + 1).ReadAsArray() for i in range(data.RasterCount)]
|
||||
im_ms = np.stack(bands, 2)
|
||||
|
||||
# cloud mask
|
||||
im_qa = im_ms[:,:,5]
|
||||
cloud_mask = sds.create_cloud_mask(im_qa, satname, plot_bool)
|
||||
cloud_mask = transform.resize(cloud_mask, (im_pan.shape[0], im_pan.shape[1]),
|
||||
order=0, preserve_range=True,
|
||||
mode='constant').astype('bool_')
|
||||
# resize the image using bilinear interpolation (order 1)
|
||||
im_ms = transform.resize(im_ms,(im_pan.shape[0], im_pan.shape[1]),
|
||||
order=1, preserve_range=True, mode='constant')
|
||||
|
||||
# check if -inf or nan values and add to cloud mask
|
||||
im_inf = np.isin(im_ms[:,:,0], -np.inf)
|
||||
im_nan = np.isnan(im_ms[:,:,0])
|
||||
cloud_mask = np.logical_or(np.logical_or(cloud_mask, im_inf), im_nan)
|
||||
|
||||
# calculate cloud cover and skip image if too high
|
||||
cloud_cover = sum(sum(cloud_mask.astype(int)))/(cloud_mask.shape[0]*cloud_mask.shape[1])
|
||||
if cloud_cover > cloud_thresh:
|
||||
print('skip ' + str(i) + ' - cloudy (' + str(cloud_cover) + ')')
|
||||
idx_skipped.append(i)
|
||||
continue
|
||||
idx_nocloud.append(i)
|
||||
|
||||
# check if image for that date is already present
|
||||
if file_names_pan[i][len(satname)+1+len(sitename)+1:len(satname)+1+len(sitename)+1+10] in date_acquired_ts:
|
||||
|
||||
# find the index of the image that is repeated
|
||||
idx_samedate = utils.find_indices(date_acquired_ts, lambda e : e == file_names_pan[i][9:19])
|
||||
idx_samedate = idx_samedate[0]
|
||||
print('cloud cover ' + str(cloud_cover) + ' - ' + str(cloud_cover_ts[idx_samedate]))
|
||||
print('acc georef ' + str(acc_georef_sorted[i]) + ' - ' + str(acc_georef_ts[idx_samedate]))
|
||||
|
||||
# keep image with less cloud cover or best georeferencing accuracy
|
||||
if cloud_cover < cloud_cover_ts[idx_samedate] - 0.01:
|
||||
skip = False
|
||||
elif acc_georef_sorted[i] < acc_georef_ts[idx_samedate]:
|
||||
skip = False
|
||||
else:
|
||||
skip = True
|
||||
|
||||
if skip:
|
||||
print('skip ' + str(i) + ' - repeated')
|
||||
idx_skipped.append(i)
|
||||
continue
|
||||
else:
|
||||
del shorelines[idx_samedate]
|
||||
del t[idx_samedate]
|
||||
del cloud_cover_ts[idx_samedate]
|
||||
del date_acquired_ts[idx_samedate]
|
||||
del acc_georef_ts[idx_samedate]
|
||||
print('keep ' + str(i) + ' - deleted ' + str(idx_samedate))
|
||||
|
||||
# pansharpen rgb image
|
||||
im_ms_ps = sds.pansharpen(im_ms[:,:,[0,1,2]], im_pan, cloud_mask, plot_bool)
|
||||
im_display = sds.rescale_image_intensity(im_ms_ps[:,:,[2,1,0]], cloud_mask, 100, False)
|
||||
# add down-sized bands for NIR and SWIR (since pansharpening is not possible)
|
||||
im_ms_ps = np.append(im_ms_ps, im_ms[:,:,[3,4]], axis=2)
|
||||
|
||||
# classify image in 4 classes (sand, whitewater, water, other) with NN classifier
|
||||
im_classif = sds.classify_image_NN(im_ms_ps, im_pan, cloud_mask, True)
|
||||
# labels
|
||||
im_sand = im_classif == 1
|
||||
im_swash = im_classif == 2
|
||||
im_water = im_classif == 3
|
||||
vec_sand = im_sand.reshape(ncols*nrows)
|
||||
vec_water = im_water.reshape(ncols*nrows)
|
||||
vec_swash = im_swash.reshape(ncols*nrows)
|
||||
|
||||
t.append(timestamps_sorted[i])
|
||||
cloud_cover_ts.append(cloud_cover)
|
||||
acc_georef_ts.append(acc_georef_sorted[i])
|
||||
date_acquired_ts.append(file_names_pan[i][9:19])
|
||||
|
||||
# calculate indices
|
||||
im_ndwi = sds.nd_index(im_ms_ps[:,:,3], im_ms_ps[:,:,1], cloud_mask, plot_bool)
|
||||
im_ndmwi = sds.nd_index(im_ms_ps[:,:,4], im_ms_ps[:,:,1], cloud_mask, plot_bool)
|
||||
im_nir = im_ms_ps[:,:,3]
|
||||
im_swir = im_ms_ps[:,:,4]
|
||||
im_ind = np.stack((im_ndwi, im_ndmwi), axis=-1)
|
||||
vec_ind = im_ind.reshape(nrows*ncols,2)
|
||||
# keep only beach
|
||||
morphology.remove_small_objects(im_sand, min_size=50, connectivity=2, in_place=True)
|
||||
# buffer around beach
|
||||
buffer_size = 7
|
||||
se = morphology.disk(buffer_size)
|
||||
im_buffer = morphology.binary_dilation(im_sand, se)
|
||||
vec_buffer = im_buffer.reshape(nrows*ncols)
|
||||
|
||||
|
||||
im = np.copy(im_display)
|
||||
im[~im_buffer,0] = 0
|
||||
im[~im_buffer,1] = 0
|
||||
im[~im_buffer,2] = 0
|
||||
plt.figure()
|
||||
plt.imshow(im)
|
||||
plt.draw()
|
||||
|
||||
int_water = vec_ind[np.logical_and(vec_buffer,vec_water),:]
|
||||
int_sand = vec_ind[np.logical_and(vec_buffer,vec_sand),:]
|
||||
int_swash = vec_ind[np.logical_and(vec_buffer,vec_swash),:]
|
||||
|
||||
fig, ax = plt.subplots(2,1, sharex=True)
|
||||
ax[0].hist(int_water[:,0], bins=100, label='water')
|
||||
ax[0].hist(int_sand[:,0], bins=100, alpha=0.5, label='sand')
|
||||
ax[0].hist(int_swash[:,0], bins=100, alpha=0.5, label='swash')
|
||||
ax[0].legend()
|
||||
ax[0].set_title('Water Index NIR-G')
|
||||
ax[1].hist(int_water[:,1], bins=100, label='water')
|
||||
ax[1].hist(int_sand[:,1], bins=100, alpha=0.5, label='sand')
|
||||
ax[1].hist(int_swash[:,1], bins=100, alpha=0.5, label='swash')
|
||||
ax[1].legend()
|
||||
ax[1].set_title('Modified Water Index SWIR-G')
|
||||
plt.draw()
|
||||
|
||||
int_all = np.append(int_water,int_sand, axis=0)
|
||||
t1 = filters.threshold_otsu(int_all[:,0])
|
||||
t2 = filters.threshold_otsu(int_all[:,1])
|
||||
|
||||
contours1 = measure.find_contours(im_ndwi, t1)
|
||||
contours2 = measure.find_contours(im_ndmwi, t1)
|
||||
|
||||
plt.figure()
|
||||
plt.imshow(im_display)
|
||||
for i,contour in enumerate(contours1): plt.plot(contour[:, 1], contour[:, 0], linewidth=3, color='c')
|
||||
for i,contour in enumerate(contours2): plt.plot(contour[:, 1], contour[:, 0], linewidth=3, color='m')
|
||||
plt.draw()
|
||||
|
||||
|
||||
plt.figure()
|
||||
ax1 = plt.subplot(1,5,1)
|
||||
plt.imshow(im_display)
|
||||
plt.xticks([])
|
||||
plt.yticks([])
|
||||
plt.axis('off')
|
||||
plt.title('RGB')
|
||||
plt.subplot(1,5,2, sharex=ax1, sharey=ax1)
|
||||
plt.imshow(im_ndwi, cmap='seismic')
|
||||
plt.xticks([])
|
||||
plt.yticks([])
|
||||
plt.axis('off')
|
||||
plt.title('NDWI')
|
||||
plt.subplot(1,5,3, sharex=ax1, sharey=ax1)
|
||||
plt.imshow(im_ndmwi, cmap='seismic')
|
||||
plt.xticks([])
|
||||
plt.yticks([])
|
||||
plt.axis('off')
|
||||
plt.title('NDMWI')
|
||||
plt.subplot(1,5,4, sharex=ax1, sharey=ax1)
|
||||
plt.imshow(im_nir, cmap='seismic')
|
||||
plt.xticks([])
|
||||
plt.yticks([])
|
||||
plt.axis('off')
|
||||
plt.title('NIR')
|
||||
plt.subplot(1,5,5, sharex=ax1, sharey=ax1)
|
||||
plt.imshow(im_swir, cmap='seismic')
|
||||
plt.xticks([])
|
||||
plt.yticks([])
|
||||
plt.axis('off')
|
||||
plt.title('SWIR')
|
||||
|
||||
fig, (ax1,ax2,ax3,ax4) = plt.subplots(4,2, figsize = (8,6))
|
||||
ax1[0].set_title('Probability density function')
|
||||
ax1[1].set_title('Cumulative distribution')
|
||||
|
||||
im = im_ndwi
|
||||
t1 = filters.threshold_otsu(im)
|
||||
vals = ax1[0].hist(im.reshape(nrows*ncols), bins=300)
|
||||
ax1[0].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
vals = ax1[1].hist(im.reshape(nrows*ncols), bins=300, cumulative=True, histtype='step')
|
||||
ax1[1].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
ax1[0].set_ylabel('NDWI')
|
||||
|
||||
im = im_ndmwi
|
||||
t1 = filters.threshold_otsu(im)
|
||||
vals = ax2[0].hist(im.reshape(nrows*ncols), bins=300)
|
||||
ax2[0].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
vals = ax2[1].hist(im.reshape(nrows*ncols), bins=300, cumulative=True, histtype='step')
|
||||
ax2[1].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
ax2[0].set_ylabel('NDMWI')
|
||||
|
||||
im = im_nir
|
||||
t1 = filters.threshold_otsu(im)
|
||||
vals = ax3[0].hist(im.reshape(nrows*ncols), bins=300)
|
||||
ax3[0].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
vals = ax3[1].hist(im.reshape(nrows*ncols), bins=300, cumulative=True, histtype='step')
|
||||
ax3[1].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
ax3[0].set_ylabel('NIR')
|
||||
|
||||
im = im_swir
|
||||
t1 = filters.threshold_otsu(im)
|
||||
vals = ax4[0].hist(im.reshape(nrows*ncols), bins=300)
|
||||
ax4[0].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
vals = ax4[1].hist(im.reshape(nrows*ncols), bins=300, cumulative=True, histtype='step')
|
||||
ax4[1].plot([t1, t1],[0, np.max(vals[0])], 'r-', label='Otsu threshold')
|
||||
ax4[0].set_ylabel('SWIR')
|
||||
|
||||
plt.draw()
|
||||
|
||||
|
||||
|
||||
#%%
|
||||
|
||||
|
||||
# calculate NDWI
|
||||
im_ndwi = sds.nd_index(im_ms_ps[:,:,3], im_ms_ps[:,:,1], cloud_mask, plot_bool)
|
||||
# detect edges
|
||||
wl_pix = sds.find_wl_contours(im_ndwi, cloud_mask, min_contour_points, True)
|
||||
# convert from pixels to world coordinates
|
||||
wl_coords = sds.convert_pix2world(wl_pix, georef)
|
||||
# convert to output epsg spatial reference
|
||||
wl = sds.convert_epsg(wl_coords, input_epsg, output_epsg)
|
||||
# classify sand pixels
|
||||
im_sand = sds.classify_sand_unsupervised(im_ms_ps, im_pan, cloud_mask, wl_pix, False, min_beach_size, plot_bool)
|
||||
|
||||
# plot a figure to select the correct water line and discard cloudy images
|
||||
plt.figure()
|
||||
cmap = cm.get_cmap('jet')
|
||||
plt.subplot(121)
|
||||
plt.imshow(im_ms_ps[:,:,[2,1,0]])
|
||||
for j,contour in enumerate(wl_pix):
|
||||
colours = cmap(np.linspace(0, 1, num=len(wl_pix)))
|
||||
plt.plot(contour[:, 1], contour[:, 0], linewidth=2, color=colours[j,:])
|
||||
plt.axis('image')
|
||||
plt.title(file_names_pan[i])
|
||||
plt.subplot(122)
|
||||
centroids = []
|
||||
for j,contour in enumerate(wl):
|
||||
colours = cmap(np.linspace(0, 1, num=len(wl)))
|
||||
centroids.append([np.mean(contour[:, 0]),np.mean(contour[:, 1])])
|
||||
plt.plot(contour[:, 0], contour[:, 1], linewidth=2, color=colours[j,:])
|
||||
plt.plot(np.mean(contour[:, 0]), np.mean(contour[:, 1]), 'o', color=colours[j,:])
|
||||
plt.plot(refpoints[:,0], refpoints[:,1], 'k.')
|
||||
plt.axis('equal')
|
||||
plt.title(file_names_pan[i])
|
||||
mng = plt.get_current_fig_manager()
|
||||
mng.window.showMaximized()
|
||||
plt.tight_layout()
|
||||
plt.draw()
|
||||
# click on the left image to discard, otherwise on the closest centroid in the right image
|
||||
pt_in = np.array(ginput(n=1, timeout=1000))
|
||||
if pt_in[0][0] < 10000:
|
||||
print('skip ' + str(i) + ' - manual')
|
||||
idx_skipped.append(i)
|
||||
continue
|
||||
# get contour that was selected (click closest to centroid)
|
||||
dist_centroid = [np.linalg.norm(_ - pt_in) for _ in centroids]
|
||||
shorelines.append(wl[np.argmin(dist_centroid)])
|
||||
|
||||
|
||||
|
||||
|
||||
# plot all shorelines
|
||||
plt.figure()
|
||||
plt.axis('equal')
|
||||
for j in range(len(shorelines)):
|
||||
plt.plot(shorelines[j][:,0], shorelines[j][:,1])
|
||||
plt.draw()
|
||||
|
||||
output = {'t':t, 'shorelines':shorelines, 'cloud_cover':cloud_cover_ts, 'acc_georef':acc_georef_ts}
|
||||
|
||||
#with open(os.path.join(filepath, sitename + '_output' + '.pkl'), 'wb') as f:
|
||||
# pickle.dump(output, f)
|
||||
#
|
||||
#with open(os.path.join(filepath, sitename + '_skipped' + '.pkl'), 'wb') as f:
|
||||
# pickle.dump(idx_skipped, f)
|
||||
#
|
||||
#with open(os.path.join(filepath, sitename + '_idxnocloud' + '.pkl'), 'wb') as f:
|
||||
# pickle.dump(idx_nocloud, f)
|
||||
Binary file not shown.
Loading…
Reference in New Issue