-
Notifications
You must be signed in to change notification settings - Fork 0
/
sensors_id.py
178 lines (156 loc) · 7.28 KB
/
sensors_id.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Ripped from sensors.py
Added a couple of different scatter plots based on the same sensor database
"""
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.patches as patches
from matplotlib import rc
import pandas as pd
scaler=2.5
rc('figure', figsize=(5.5,3.5))
rc('legend', fontsize='small')
#rc('figure', grid=True)
rc('font', family='serif')
rc('font', size=9)
rc('lines', markersize=3.0)
def LoadSensorData():
"""
Just read in the data and calculate some derived metrics from Jonny's
sheet of sensor properties
"""
sensors = pd.read_csv('figures/sensors.csv')
# JD Note: changed this to match my updated KPI which doesn't put p_px in
# denominator because I don't think that made sense.
sensors['kpi1'] = sensors['FWC'] * sensors['Width'] * sensors['Height'] * \
sensors['FPS']# / \
# sensors['Pixel Size']
bits = 12 # Could use log2(DR) but why penalize sensors with good noise characteristics?
# Data intensity in bits/sec
sensors['I_d'] = sensors['Width'] * sensors['Height'] * sensors['FPS'] * bits
# Number of samples needed for SNR=100 at alpha=5
sensors['n_samp'] = np.ceil(5e4/sensors['FWC'])
# Number of samples for each ground point for GSD=1m at h=500km (v=7000m/s)
sensors['n_samp_gsd1'] = np.floor(sensors['Height']/(7000./1./sensors['FPS']))
# Number of samples for each ground point for GSD=4m at h=500km (v=7000m/s)
sensors['n_samp_gsd4'] = np.floor(sensors['Height']/(7000./4./sensors['FPS']))
# Number of bands that can be sampled at SNR=100 with GSD=1m. Not floored/rounded...
sensors['n_bands_gsd1'] = sensors['n_samp_gsd1']/sensors['n_samp']
# Number of bands that can be sampled at SNR=100 with GSD=1m. Not floored/rounded...
sensors['n_bands_gsd4'] = sensors['n_samp_gsd4']/sensors['n_samp']
# Data intensity assuming on-board summing with shift/add
# Could multiply by N_bands below but that's less fundamental
sensors['I_d_sum'] = sensors['I_d']/sensors['n_samp_gsd1']
return(sensors)
def SensorType(sensors):
ccds = sensors.loc[sensors['Type'] == 'CCD']
cmos = sensors.loc[(sensors['Type'] == 'CMOS') & (sensors['Shutter'] == 'Global')]
cmos_rolling = sensors.loc[(sensors['Type'] == 'CMOS') & (sensors['Shutter'] == 'Rolling')]
return(ccds, cmos, cmos_rolling)
def PlotPsiVsId(sensors):
"""
Make some plots of
"""
ccds, cmos, cmos_rolling = SensorType(sensors)
# Information content vs data intensity. Scatter about the linear fit is just
# due to higher order variation in e.g. aspect ratio, bit depth, etc.
fig1 = plt.figure(1)
plt.loglog(ccds['I_d'], ccds['kpi1'], 'r^', label='CCD')
plt.loglog(cmos['I_d'], cmos['kpi1'], 'go', label='CMOS (Global)')
plt.loglog(cmos_rolling['I_d'], cmos_rolling['kpi1'], 'bs', label='CMOS (Rolling)')
plt.xlabel(r'$BW_{det}$ [bits/sec]')
plt.ylabel(r'$\psi_{px} = h_{px}N_{e^-}^{FWC} LR$')
#plt.xlim(2, 8)
#plt.ylim(0, 50)
plt.tight_layout()
plt.legend(loc='best')
# Information content vs data intensity assuming on-board summing of digital
# oversamples. This (of course) favors CMOS detectors, shifting them to the
# left so that they are above CCDs instead of up and to the right.
fig2 = plt.figure(2)
plt.loglog(ccds['I_d_sum'], ccds['kpi1'], 'r^', label='CCD')
plt.loglog(cmos['I_d_sum'], cmos['kpi1'], 'go', label='CMOS (Global)')
plt.loglog(cmos_rolling['I_d_sum'], cmos_rolling['kpi1'], 'bs', label='CMOS (Rolling)')
plt.xlabel(r'$BW_{det}$ [bits/sec]')
plt.ylabel(r'$\psi_{px} = h_{px}N_{e^-}^{FWC} LR$')
#plt.xlim(2, 8)
#plt.ylim(0, 50)
plt.tight_layout()
plt.legend(loc='best')
return(fig1, fig2)
def PlotNBands(sensors):
# Klugy attempt to show how the height of a CMOS detector can be used to
# add more spectral bands
font = {'family': 'sans-serif',
'color': 'darkred',
'weight': 'normal',
'size': 9,
}
# Some interpolated grids for plotting the number of bands on a non-linear
# Y axis
yax_grid = np.array([0,0.5,0.6,1,1.5,1.6,2,3,4,5,6,7,10,15,20,25,30])
sensors['yax'] = np.interp(sensors['n_bands_gsd1'], yax_grid, np.arange(len(yax_grid)))
yax_grid4 = np.array([0,0.5,0.6,1,1.5,1.6,2,3,4,5,6,7,10,15,20,25,30,50])
sensors['yax4'] = np.interp(sensors['n_bands_gsd4'], yax_grid4, np.arange(len(yax_grid4)))
ccds, cmos, cmos_rolling = SensorType(sensors)
fig1,ax = plt.subplots()
ax.plot(ccds['Pixel Size'], ccds['yax'], 'r^', label='CCD')
ax.plot(cmos['Pixel Size'], cmos['yax'], 'go', label='CMOS (Global)')
ax.plot(cmos_rolling['Pixel Size'], cmos_rolling['yax'], 'bs', label='CMOS (Rolling)')
ax.set_xlabel(r'$p_{px}$ ($\mu$m)')
ax.set_ylabel(r'$N_{bands}$')
tickind = [0,3,6,7,8,9,12,15]
ax.set_yticks(tickind)
ax.set_yticklabels(np.array(yax_grid[tickind]))
ax.set_xlim(2, 8)
ax.set_title("Channels available for GSD=1m, SNR=100")
plt.tight_layout()
ax.legend(loc='best')
labelx = 2.2
ax.add_patch(patches.Rectangle((2,-1),6,4,facecolor='grey', alpha=0.2))
ax.text(labelx, 0.2, r"$SNR<100$",fontdict=font)
ax.add_patch(patches.Rectangle((2,3),6,4,facecolor='grey', alpha=0.1))
ax.text(labelx, 4, "Panchromatic",fontdict=font)
#ax.add_patch(patches.Rectangle((2,6),6,3,facecolor='red', alpha=0.1))
#ax.text(6.8, 7, "False color",fontdict=font)
ax.add_patch(patches.Rectangle((2,7),6,5,facecolor='green', alpha=0.1))
ax.text(labelx, 9.5, "Multispectral",fontdict=font)
ax.add_patch(patches.Rectangle((2,12),6,5,facecolor='blue', alpha=0.1))
ax.text(labelx, 13, "Hyperspectral",fontdict=font)
#f.savefig('figures/p_kpi.pgf')
fig2,ax = plt.subplots()
ax.plot(ccds['Pixel Size'], ccds['yax4'], 'r^', label='CCD')
ax.plot(cmos['Pixel Size'], cmos['yax4'], 'go', label='CMOS (Global)')
ax.plot(cmos_rolling['Pixel Size'], cmos_rolling['yax4'], 'bs', label='CMOS (Rolling)')
ax.set_xlabel(r'$p_{px}$ ($\mu$m)')
ax.set_ylabel(r'$N_{bands}$')
tickind = [0,3,6,7,8,9,12,15]
ax.set_yticks(tickind)
ax.set_yticklabels(np.array(yax_grid4[tickind]))
ax.set_xlim(2, 8)
ax.set_title("Channels available for GSD=4m, SNR=100")
plt.tight_layout()
ax.legend(loc='best')
labelx = 2.2
ax.add_patch(patches.Rectangle((2,-1),6,4,facecolor='grey', alpha=0.2))
ax.text(labelx, 0.8, r"$SNR<100$",fontdict=font)
ax.add_patch(patches.Rectangle((2,3),6,4,facecolor='grey', alpha=0.1))
ax.text(labelx, 4, "Panchromatic",fontdict=font)
#ax.add_patch(patches.Rectangle((2,6),6,3,facecolor='red', alpha=0.1))
#ax.text(6.8, 7, "False color",fontdict=font)
ax.add_patch(patches.Rectangle((2,7),6,5,facecolor='green', alpha=0.1))
ax.text(labelx, 10, "Multispectral",fontdict=font)
ax.add_patch(patches.Rectangle((2,12),6,6,facecolor='blue', alpha=0.1))
ax.text(labelx, 16, "Hyperspectral",fontdict=font)
return(fig1,fig2)
if __name__ == "__main__":
sensors = LoadSensorData()
fig1,fig2 = PlotPsiVsId(sensors)
fig3,fig4 = PlotNBands(sensors)
fig1.savefig("figures/psi_vs_id.pgf")
fig2.savefig("figures/psi_vs_id_summed.pgf")
fig3.savefig("figures/nbands_1m.pgf")
fig4.savefig("figures/nbands_4m.pgf")
plt.show()