-
Notifications
You must be signed in to change notification settings - Fork 141
/
synimagegen.py
305 lines (234 loc) · 10.8 KB
/
synimagegen.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
"""This module contains a pure python implementation of the basic
cross-correlation algorithm for PIV image processing."""
__licence_ = """
Copyright (C) 2011 www.openpiv.net
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
"""
import os
import numpy as np
import scipy
import scipy.special
import scipy.interpolate
from scipy import io
import matplotlib.pyplot as pl
import matplotlib.cm as cm
from PIL import Image
class continuous_flow_field:
def __init__(self,data,inter=False):
'''
Checks if the continous flow should be created from a set of data points
if so it interpolates them for a continuous flow field
'''
self.inter = inter
if inter:
self.f_U = scipy.interpolate.interp2d(data[:,0],data[:,1],data[:,2])
self.f_V = scipy.interpolate.interp2d(data[:,0],data[:,1],data[:,3])
'''
Defining a synthetic flow field
'''
def f_U(self,x,y):
#example for synthetic U velocity
u=2.5+0.5*np.sin((x**2+y**2)/0.01)
return u
def f_V(self,x,y):
#example for synthetic V velocity
v=0.5+0.1*np.cos((x**2+y**2)/0.01)
return v
def get_U_V(self,x,y):
#return the U and V velocity at a certain position
if self.inter:
return self.f_U(x,y)[0],self.f_V(x,y)[0]
else:
return self.f_U(x,y),self.f_V(x,y)
def create_syn_quiver(self,number_of_grid_points,path=None):
#return and save a synthetic flow map
X,Y = np.meshgrid(np.linspace(0,1,number_of_grid_points),np.linspace(0,1,number_of_grid_points))
U = np.zeros(X.shape)
V = np.zeros(Y.shape)
for r in range(X.shape[0]):
for c in range(X.shape[1]):
u,v = self.get_U_V(X[r,c],Y[r,c])
U[r,c] = u
V[r,c] = v
m = np.sqrt(np.power(U, 2) + np.power(V, 2))
fig = pl.quiver(X, Y, U, V,m,clim=[1.5,m.max()],scale=100,width=0.002,headwidth=6,minshaft=2)
cb = pl.colorbar(fig)
cb.set_clim(vmin=1.5, vmax=m.max())
if not path:
pl.savefig('syn_quiver.png', dpi=400)
pl.close()
else:
pl.savefig(path + 'syn_quiver.png', dpi=400)
pl.close()
return X,Y,U,V
def create_synimage_parameters(input_data,x_bound,y_bound,image_size,path='None',inter=False,den=0.008,per_loss_pairs=2,par_diam_mean=15**(1.0/2),par_diam_std=1.5,par_int_std=0.25,dt=0.1):
"""Creates the synthetic image with the synthetic image parameters
Parameters
----------
input_data: None or numpy array
If you have data from which to genrate the flow feild the synthetic image.
It should be passed on as a numpy array with columns being (X grid position,Y grid position,U velocity at (X,Y) grid point,V velocity at (X,Y) grid point)
Else, pass None and define a synthetic flow field in continuous_flow_field class.
x_bound,y_bound: list/tuple of floats
The boundries of interest in the synthetic flow field.
image_size: list/tuple of ints
The desired image size in pixels.
path: str('None' for no generating data)
Path to txt file of input data.
inter: boolean
False if no interpolation of input data is needed.
True if there is data you want to interpolate from.
den: float
Defines the number of particles per image.
per_loss_pairs: float
Percentage of synthetic pairs loss.
par_diam_mean: float
Mean particle diamter in pixels.
par_diam_std: float
Standard deviation of particles diamter in pixels.
par_int_std: float
Standard deviation of particles intensities.
dt: float
Synthetic time difference between both images.
Returns
-------
ground_truth: continuous_flow_field class
The synthetic ground truth as a continuous_flow_field class.
cv:
Convertion value to convert U,V from pixels/images to meters/seconds.
x_1,y_1: numpy array
Position of particles in the first synthetic image.
U_par,V_par: numpy array
Velocity speeds for each particle.
par_diam1: numpy array
Particle diamters for the first synthetic image.
par_int1: numpy array
Particle intensities for the first synthetic image.
x_2,y_2: numpy array
Position of particles in the second synthetic image.
par_diam2: numpy array
Particle diamters for the second synthetic image.
par_int2: numpy array
Particle intensities for the second synthetic image.
"""
#Data processing
if not path == 'None':
f = open(path,'r')
data = f.readlines()
f.close()
data = [line.split('\t') for line in data]
data = np.array(data).astype(float)
data = np.array([line for line in data.tolist() if 1.2*x_bound[1]>=line[1]>=0.8*x_bound[0] and 1.2*y_bound[1]>=line[2]>=0.8*y_bound[0]])
else:
data = input_data
if inter:
cff = continuous_flow_field(data,inter=True)
else:
cff = continuous_flow_field(None)
#Creating syn particles
num_of_par = int(image_size[0]*image_size[1]*den)
num_of_lost_pairs = num_of_par*(per_loss_pairs/100)
x_1 = np.random.uniform(x_bound[0]*0.8,x_bound[1]*1.2,num_of_par)
y_1 = np.random.uniform(y_bound[0]*0.8,y_bound[1]*1.2,num_of_par)
par_diam1 = np.random.normal(par_diam_mean,par_diam_std,num_of_par)
particleCenters = np.random.uniform(size=num_of_par)-0.5
par_int1 = np.exp(-particleCenters**2/(2*par_int_std**2))
U_par = np.zeros(x_1.shape)
V_par = np.zeros(y_1.shape)
x_2 = np.zeros(x_1.shape)
y_2 = np.zeros(y_1.shape)
par_diam2 = np.zeros(par_diam1.shape)
par_int2 = np.zeros(par_int1.shape)
def Move_par(i):
U_par[i],V_par[i] = cff.get_U_V(x_1[i],y_1[i])
x_2[i] = x_1[i]+U_par[i]*dt
y_2[i] = y_1[i]+V_par[i]*dt
par_diam2[i] = par_diam1[i]
par_int2[i] = par_int1[i]
cpl = 0
for i in range(num_of_par):
if cpl<num_of_lost_pairs:
if -0.4>particleCenters[i] or 0.4<particleCenters[i]:
per_to_lose = 1-(0.5 - np.abs(particleCenters[i]))/0.1
if np.random.uniform()<min(per_loss_pairs/10,1)*per_to_lose:
x_2[i] = np.random.uniform(x_bound[0]*0.8,x_bound[1]*1.2)
y_2[i] = np.random.uniform(y_bound[0]*0.8,y_bound[1]*1.2)
par_diam2[i] = np.random.normal(par_diam_mean,par_diam_std)
par_int2[i] = np.exp(-(np.random.uniform()-0.5)**2/(2*par_int_std**2))
cpl+=1
else:
Move_par(i)
else:
Move_par(i)
else:
Move_par(i)
print('Requested pair loss:',str(int(num_of_lost_pairs)),' Actual pair loss:',str(cpl))
xy_1 = np.transpose(np.vstack((x_1,y_1,U_par,V_par,par_diam1,par_int1)))
xy_2 = np.transpose(np.vstack((x_2,y_2,par_diam2,par_int2)))
#Choosing particles in boundary area
bounded_xy_1 = np.asarray([xy for xy in xy_1 if x_bound[1]>=xy[0]>=x_bound[0] and y_bound[1]>=xy[1]>=y_bound[0]])
bounded_xy_2 = np.asarray([xy for xy in xy_2 if x_bound[1]>=xy[0]>=x_bound[0] and y_bound[1]>=xy[1]>=y_bound[0]])
#Tranforming coordinates into pixels
x1 = ((bounded_xy_1[:,0]-x_bound[0])/(x_bound[1]-x_bound[0]))*image_size[0]
y1 = ((bounded_xy_1[:,1]-y_bound[0])/(y_bound[1]-y_bound[0]))*image_size[1]
x2 = ((bounded_xy_2[:,0]-x_bound[0])/(x_bound[1]-x_bound[0]))*image_size[0]
y2 = ((bounded_xy_2[:,1]-y_bound[0])/(y_bound[1]-y_bound[0]))*image_size[1]
conversion_value = min((x_bound[1]-x_bound[0])/image_size[0],(y_bound[1]-y_bound[0])/image_size[1])/dt
return cff,conversion_value,x1,y1,bounded_xy_1[:,2],bounded_xy_1[:,3],bounded_xy_1[:,4],bounded_xy_1[:,5],x2,y2,bounded_xy_2[:,2],bounded_xy_2[:,3]
def generate_particle_image(HEIGHT, WIDTH, X, Y, PARTICLE_DIAMETERS, PARTICLE_MAX_INTENSITIES,BIT_DEPTH):
"""Creates the synthetic image with the synthetic image parameters
Should be run with the parameters of each image (first,second) separately.
Parameters
----------
HEIGHT, WIDTH: int
The number of pixels in the desired output image.
X,Y: numpy array
The X and Y positions of the particles, created by create_synimage_parameters().
PARTICLE_DIAMETERS, PARTICLE_MAX_INTENSITIES: numpy array
The intensities and diameters of the particles, created by create_synimage_parameters().
BIT_DEPTH: int
The bit depth of the desired output image.
Returns
-------
Image: numpy array
The desired synthetic image.
"""
render_fraction = 0.75
IMAGE_OUT = np.zeros([HEIGHT, WIDTH])
minRenderedCols = (X - render_fraction * PARTICLE_DIAMETERS).astype(int)
maxRenderedCols = (np.ceil(X + render_fraction * PARTICLE_DIAMETERS)).astype(int)
minRenderedRows = (Y - render_fraction * PARTICLE_DIAMETERS).astype(int)
maxRenderedRows = (np.ceil(Y + render_fraction * PARTICLE_DIAMETERS)).astype(int)
index_to_render = []
for i in range(X.size):
if 1<minRenderedCols[i] and maxRenderedCols[i]< WIDTH and 1<minRenderedRows[i] and maxRenderedRows[i]< HEIGHT:
index_to_render.append(i)
for i in range(len(index_to_render)):
ind = index_to_render[i]
max_int = PARTICLE_MAX_INTENSITIES[ind]
par_diam = PARTICLE_DIAMETERS[ind]
sqrt8 = np.sqrt(8)
x = X[ind]
y = Y[ind]
bl = max(minRenderedCols[ind],0)
br = min(maxRenderedCols[ind],WIDTH)
bu = max(minRenderedRows[ind],0)
bd = min(maxRenderedRows[ind],HEIGHT)
for c in range(bl,br):
for r in range(bu,bd):
IMAGE_OUT[r,c] = IMAGE_OUT[r,c] + max_int * par_diam**2 * np.pi / 32 * \
( scipy.special.erf( sqrt8 * (c - x - 0.5) / par_diam ) - scipy.special.erf(sqrt8 * (c - x + 0.5) / par_diam)) * \
( scipy.special.erf( sqrt8 * (r - y - 0.5)/ par_diam) - scipy.special.erf(sqrt8 *(r - y + 0.5) / par_diam))
NOISE_MEAN = 2**(BIT_DEPTH*0.3)
NOISE_STD = 0.25*NOISE_MEAN
Noise = NOISE_STD * np.random.randn(HEIGHT, WIDTH) + NOISE_MEAN
return (IMAGE_OUT*(2**BIT_DEPTH * 2.8**2/8)+Noise).astype(int)[::-1]