forked from SarveshVGharat/Plages-Identification
-
Notifications
You must be signed in to change notification settings - Fork 0
/
segment.py
474 lines (343 loc) · 14.2 KB
/
segment.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
import cv2
import math
import time
import tqdm
import numpy as np
import datetime
import random
import multiprocessing
from configs import *
import matplotlib.pyplot as plt
from google.cloud.exceptions import NotFound
from sklearn.linear_model import LinearRegression
from sklearn.metrics import r2_score
from circle_fit import *
from astropy.time import Time
#try keeping lower limit as well
#add functionality to do images in bulk
#take a look at those papers mentioned by reviewer 2
def algorithm(
input_file = None,
thresh = None,
clip_limit = None,
area_thresh = None,
input_image = None,
lower_area_thresh = None,
override = False,
get_segmented_image = False
):
date_time = return_date_and_time(file = input_file)
date, time = date_time.split(" ")
corrected_area = get_corrected_area(date = date)
if corrected_area == -99 and override == False:
return False,False,False
if input_image is None:
input_image = cv2.imread(input_file,0)
else:
input_image = input_image
crop_size = 500
padding = int((800-crop_size)/2)
input_image = center_crop(input_image, crop_size, crop_size)
input_image = cv2.copyMakeBorder(input_image,padding,padding,padding,padding,cv2.BORDER_CONSTANT,value=0)
binary_threshold = thresh
disc_contour = return_solar_circumference_contour(image = input_image)
disc_area = cv2.contourArea(disc_contour)
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(3,3))
clahe = cv2.createCLAHE(clipLimit = clip_limit) #2
clahe_image = clahe.apply(input_image)
thresh_val, thresh_img = cv2.threshold(clahe_image, binary_threshold, 255, cv2.THRESH_BINARY)
thresh_img = cv2.morphologyEx(thresh_img, cv2.MORPH_OPEN, kernel)
thresh_img = cv2.erode(thresh_img,kernel,iterations=2)
thresh_img = cv2.dilate(thresh_img,kernel,iterations=1)
new_mask_M = np.zeros( input_image.shape, dtype="uint8" )
output_image = thresh_img
disc_contour = np.squeeze(disc_contour)
disc_contour = [list(ele) for ele in disc_contour]
Xc, Yc, R, sigma = taubinSVD(disc_contour)
Xc = np.ceil(Xc)
Yc = np.ceil(Yc)
Bo, Lo, P, Rad = calc_Bo_Lo_P_Rad(date_str = date_time)
contours, h = cv2.findContours(output_image,cv2.RETR_LIST,cv2.CHAIN_APPROX_NONE)
area = 0.0
for contour in contours:
sigma_thetaIn = 0
sigma_lIn = 0
sigma_In =0
mask = np.zeros(output_image.shape, dtype="uint8" )
cv2.drawContours(mask, [contour],-1,(255,255,255),-1)
candidate_points = return_non_zero_points(image = mask)
for point in candidate_points:
i = point[0]
j = point[1]
In = mask[i,j]
x = j
y = i
r, theta_prime = convert_to_polar(x = x, y = y, Xc = Xc, Yc = Yc, R = R)
T = Rad/15
R_nought = Rad/(7*36)*(1e-12)*29.5953*np.cos(math.radians(math.degrees(np.arccos(-0.00629*T))/3+240))
rho_prime = R_nought * r/R
rho = math.degrees(np.arcsin(np.sin(math.radians(rho_prime))/np.sin(math.radians(R_nought)))) - rho_prime
sin_theta = np.cos(math.radians(rho))*np.sin(math.radians(Bo)) + np.sin(math.radians(rho))*np.cos(math.radians(Bo))*np.sin(math.radians(theta_prime))
cos_theta = np.sqrt(1-sin_theta**2)
sin_l = np.cos(math.radians(theta_prime))*np.sin(math.radians(rho))/cos_theta
theta = np.arcsin(sin_theta)
l = np.arcsin(sin_l)
sigma_thetaIn = sigma_thetaIn + theta*In
sigma_lIn = sigma_lIn + l*In
sigma_In = sigma_In + In
theta_plage = sigma_thetaIn/sigma_In
l_plage = sigma_lIn/sigma_In
cos_delta = np.sin(math.radians(Bo))*np.sin(theta_plage)+np.cos(math.radians(Bo))*np.cos(theta_plage)*np.cos(l_plage)
pixel_size = Rad/R
pixel_area = pixel_size**2
#area = area + (cv2.contourArea(contour)*pixel_area)/(2*np.pi*(Rad)**2 * cos_delta)
contour_area = (cv2.contourArea(contour))/(cos_delta)
disc_area_px = disc_area*pixel_area
if contour_area < area_thresh:
if (lower_area_thresh is not None and contour_area > lower_area_thresh) or lower_area_thresh is None:
area = area + (contour_area*pixel_area)/disc_area_px
new_mask_M = new_mask_M + mask
output_image = new_mask_M
contours, h = cv2.findContours(output_image,cv2.RETR_LIST,cv2.CHAIN_APPROX_NONE)
input_image_rgb = cv2.cvtColor(input_image, cv2.COLOR_GRAY2RGB)
cv2.drawContours(input_image_rgb, contours, -1, (0, 0, 255), 2)
#input_image_rgb = input_image = cv2.copyMakeBorder(input_image_rgb,padding,padding,padding,padding,cv2.BORDER_CONSTANT,value=0)
if not(get_segmented_image):
return corrected_area, area, input_image_rgb
else:
return corrected_area, area, output_image
def return_solar_circumference_contour(image = None):
contours, h = cv2.findContours(image,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_NONE)
contours = sorted(contours, key=cv2.contourArea, reverse=True)
return contours[0]
def center_crop(image, new_height, new_width):
height, width = image.shape[:2]
start_row = (height - new_height) // 2
start_col = (width - new_width) // 2
cropped_image = image[start_row:start_row+new_height, start_col:start_col+new_width]
return cropped_image
def get_dataframe(id = None):
# Set up the query
query = """
WITH base_query AS (
SELECT
date,
time,
corrected_area,
calculated_area
FROM
`{}`
)
SELECT
date,
STRING_AGG(time, ",") AS time,
AVG(corrected_area) AS corrected_area,
AVG(calculated_area) AS calculated_area
FROM
base_query
GROUP BY
date
ORDER BY
date ASC;
""".format(id)
query_job = client.query(query)
dataframe = (
query_job
.result()
.to_dataframe()
)
return dataframe
def plot_time_series(df = None, return_fig = False):
rolling_window_size = 30
df['corrected_area_rolling_mean'] = df['corrected_area'].rolling(rolling_window_size).mean()
df['calculated_area_rolling_mean'] = df['calculated_area'].rolling(rolling_window_size).mean()
fig = plt.figure(figsize=(10,6))
plt.plot(
df['corrected_area_rolling_mean'],
label='Corrected Plage Index {} point Rolling Mean (Chatzistergos et al., 2020)'.format(rolling_window_size)
)
plt.plot(
df['calculated_area_rolling_mean'],
label='Calculated Plage Index {} point Rolling Mean (Our approach)'.format(rolling_window_size)
)
x_axis_ticks = np.arange(100, len(df), 150)
x_axis_labels = df["date"].iloc[x_axis_ticks]
plt.xticks(x_axis_ticks, x_axis_labels)
plt.xlabel('date(yyyy-mm-dd)')
plt.ylabel('plage index(disc fraction)')
plt.legend()
if return_fig:
return fig
plt.show()
def plot_scatter_plot(df = None, return_fig = False):
fig = plt.figure(figsize=(8,6))
X = df['corrected_area'].values.reshape(-1, 1)
Y = df['calculated_area'].values.reshape(-1, 1)
reg = LinearRegression().fit(X, Y)
Y_pred = reg.predict(X)
# Calculate the R^2 score
r2 = r2_score(Y, Y_pred)
# Calculate the correlation coefficient
corr = np.corrcoef(X.reshape(-1), Y.reshape(-1))[0, 1]
plt.xlim([0.0, 0.1])
plt.ylim([0.0, 0.1])
# Plot the scatter plot
plt.scatter(X, Y)
# Plot the regression line
plt.plot(X, Y_pred, color='red')
# Add the R^2 value and the correlation coefficient to the plot
plt.text(0.08, 0.015, 'R^2 = {:.2f}'.format(r2), fontsize=10)
plt.text(0.08, 0.01, 'r = {:.2f}'.format(corr), fontsize=10)
# Set the x and y axis labels
plt.xlabel("Corrected Plage Index (Chatzistergos et al., 2020)")
plt.ylabel("Calculated Plage Index (Our approach)")
if return_fig:
return fig
plt.show()
def modify_plots():
df = get_dataframe(id = table_id)
plot_time_series(df = df)
plot_scatter_plot(df = df)
def calc_Bo_Lo_P_Rad(date_str = None):
time = Time(date_str, format = 'iso', scale = 'utc')
time.delta_ut1_utc = -19800
time = Time(time.ut1.iso, format = 'iso', scale = 'utc')
JD = time.jd
T = (JD-2415020)/36525
L_prime = 279.69668 + 36000.76892*T + 0.0003025*(T**2)
g = 358.47583 + 35999.04975*T - 0.00015*(T**2) - 0.0000033*(T**3)
ohm = 259.18 - 1934.142*T
C1 = (1.91946 - 0.004789*T - 0.000014*(T**2))*np.sin(math.radians(g))
C2 = (0.020094 - 0.0001*T)* np.sin(math.radians(2*g))
C3 = 0.0001*np.sin(math.radians(3*g))
C = C1+C2+C3
Lambda_nought = L_prime + C
Lambda_a = Lambda_nought - 0.00569 - 0.00479* np.sin(math.radians(ohm))
phi = (360/25.38)*(JD - 2398220)
if not(phi>=0 and phi<=360):
while not(phi>=0 and phi<=360):
if phi>360:
phi = phi-360
else:
phi = phi+360
#scale phi to zero to 360
K = 74.3646 + 1.395833*T
I = 7.25
epsilon_nought = 23.452295 - 0.0130125*T - 0.00000164*(T**2) + 0.000000593*(T**3)
epsilon = epsilon_nought + 0.00256*np.cos(math.radians(ohm))
X = np.arctan(-1*np.cos(math.radians(Lambda_a))*np.tan(math.radians(epsilon)))
Y = np.arctan(-1* np.cos(math.radians(Lambda_nought - K))*np.tan(math.radians(I)))
P = math.degrees(X+Y)
Bo = math.degrees(np.arcsin(np.sin(math.radians(Lambda_nought - K))*np.sin(math.radians(I))))
M = 360 - phi
Lo = math.degrees(np.arctan((np.sin(math.radians(K-Lambda_nought))*np.cos(math.radians(I)))/(-1* np.cos(math.radians(K-Lambda_nought))))) + M
R_prime = 1.00014 - 0.01671*np.cos(math.radians(g)) - 0.00014 * np.cos(math.radians(2*g))
Rad = 0.2666/R_prime * 3600
return Bo, Lo, P, Rad
#20230206070710
def convert_to_polar(x = None, y = None, Xc = None, Yc= None, R = None):
x = x - Xc
y = y - Yc
theta_prime = math.degrees(np.arctan2(y,x))
r = np.clip(np.sqrt(x**2+y**2), a_max = R-1, a_min = 0)
return r, theta_prime
def return_non_zero_points(image = None):
non_zero_indices = np.nonzero(image)
non_zero_indices = zip(non_zero_indices[0], non_zero_indices[1])
return non_zero_indices
def return_date_and_time(file = None):
date_time = file.split("_")[1].replace("T", " ")
dt = datetime.datetime.strptime(date_time, "%Y%m%d %H%M%S")
formatted_dt = dt.strftime("%Y-%m-%d %H:%M:%S")
return formatted_dt
def get_corrected_area(date = None):
year, month, day = map(int, date.split("-"))
row = df.loc[(df["Year"] == year) & (df["Month"] == month) & (df["Day"] == day)]
corrected_area = row["Corrected Area"].values[0]
return corrected_area
def create_table(id = None):
schema = [
bigquery.SchemaField("date", "STRING", mode="NULLABLE"),
bigquery.SchemaField("time", "STRING", mode="NULLABLE"),
bigquery.SchemaField("corrected_area", "FLOAT", mode="NULLABLE"),
bigquery.SchemaField("calculated_area", "FLOAT", mode="NULLABLE"),
bigquery.SchemaField("thresh_manual", "INTEGER", mode="NULLABLE")
]
table = bigquery.Table(id, schema=schema)
table = client.create_table(table)
def delete_table(id = None):
client.delete_table(id, not_found_ok=False)
def delete_table_rows(id = None):
query = """
DELETE FROM `{}` WHERE true;
""".format(id)
query_job = client.query(query)
def append_to_table(id = None, date = None, time = None, corrected_area = None, calculated_area = None, thresh_manual = None):
rows_to_insert = [
{u'date':date, u'time':time, u'corrected_area':corrected_area, u'calculated_area':calculated_area, u'thresh_manual':thresh_manual },
]
client.insert_rows_json(id, rows_to_insert)
def pipeline(file = None):
file_name = file.split('/')[-1]
thresh_list = [100, 120, 140, 150, 160, 170, 180, 190, 200, 220]
calculated_area = None
corrected_area = None
thresh_manual = None
diff = None
for thresh in thresh_list:
corrected_area, calculated_area_temp, output_image_temp = algorithm(
input_file = file,
thresh = thresh,
clip_limit=clip_limit_value,
area_thresh = area_thresh_value,
lower_area_thresh=lower_area_thresh_value
)
if not(corrected_area == False):
thresh_manual = thresh
if calculated_area is None:
calculated_area = calculated_area_temp
output_image = output_image_temp
diff = np.abs(corrected_area - calculated_area)
else:
diff_temp = np.abs(corrected_area - calculated_area_temp)
if diff_temp < diff:
calculated_area = calculated_area_temp
output_image = output_image_temp
diff = diff_temp
else:
break
formatted_dt = return_date_and_time(file)
date, time = formatted_dt.split(" ")
if not(corrected_area == False):
append_to_table(
id = table_id,
date=date,
time=time,
corrected_area=corrected_area,
calculated_area=calculated_area,
thresh_manual=thresh_manual
)
cv2.imwrite(filename = output_image_dir + '/' + file_name, img = output_image)
def main():
try:
client.get_table(table_id)
print("Table {} already exists.Deleting table and recreating.".format(table_id))
time.sleep(60)
delete_table_rows(id = table_id)
time.sleep(60)
except NotFound:
print("Table {} is not found.Creating table.".format(table_id))
time.sleep(60)
create_table(id = table_id)
time.sleep(60)
os.makedirs(output_image_dir, exist_ok=True)
files = os.listdir(images_dir)
input_files = [images_dir + '/'+ file for file in files]
pool = multiprocessing.Pool(processes = 12)
for _ in tqdm.tqdm(pool.imap_unordered(pipeline, input_files), total=len(input_files)):
pass
df = get_dataframe(id = table_id)
plot_time_series(df = df)
plot_scatter_plot(df = df)
if __name__ == '__main__':
main()