def analisisHR(hrValues, miliValues, sample_size):
plt.plot(miliValues, hrValues)
#plt.show();
smoothed_values = [
data for data, i in smooth_curve_simple(hrValues, sample_size)
]
average = []
peaks = fp(smoothed_values)[0]
previous = peaks[0]
peakplotX = [miliValues[previous]]
peakplotY = [smoothed_values[previous]]
for id in peaks[1:]:
average.append(miliValues[id] - miliValues[previous])
previous = id
peakplotX.append(miliValues[previous])
peakplotY.append(smoothed_values[previous])
plt.plot(smoothed_values)
#plt.scatter(peakplotY , marker='+', c='Red')
#plt.show()
print("Picos: ", len(peaks), " Longitud: ", miliValues[-1] - miliValues[0])
len(peaks)
valorHR = (60000) / ((miliValues[-1] - miliValues[0]) / len(peaks))
return valorHR
def peak_detection(y,show = False): '''Function which detects te number of peaks in a curve, according to a certain threshold. First peak must be 1/2 of the max and second peak must be 1/2 or more of the main peak, with a 1/4 minimum prominence''' from scipy.signal import find_peaks as fp ydet = np.concatenate((y[-12:],y,y[:12])) # we repeat the curve a little bit such that peaks at borders are not neglected by algo p, prop = fp( ydet, height=0, distance=4, prominence = (np.amax(y)/6,y.max())) idx = ((p-12 > -1) & (p-12 < 24)) p = (p - 12)[idx] for key in prop.keys(): prop[key] = prop[key][idx] return p, prop
def analisisHR2(hrValues, miliValues, sample_size):
#smoothed_values = smooth_curve_simple(hrValues, sample_size)
smoothed_values = [
data for data, i in smooth_curve_simple(hrValues, sample_size)
]
peaks = fp(smoothed_values)[0]
average = []
for id in range(0, len(peaks) - 1, 2):
current = peaks[id + 1] * sample_size + sample_size
previous = peaks[id] * sample_size + sample_size
average.append(miliValues[current] - miliValues[previous])
# previous = id*sample_size
plt.plot(smoothed_values)
plt.scatter(peaks, [smoothed_values[j] for j in peaks],
marker='+',
c='Red')
#plt.show()
valorHR = (60000 * len(peaks)) / (miliValues[-1] - miliValues[0])
# valorHR=(60000)/(sum(average)/len(average))
return valorHR
def find_peaks(points, threshold=0.5):
"""
Finds the peaks in a signal.
Parameters
----------
points : ndarray
1-D array of normalized data points.
threshold : float
Limit to filter peaks; those below that value are discarded.
Returns
-------
ndarray : 1-D array with the indices corresponding to the peaks of
points.
"""
# Find the peaks
peaks, _ = fp(points, distance=15)
# Filter the peaks based on threshold
valid_peaks = np.argwhere(points[peaks] > threshold)
return peaks[valid_peaks.ravel()]
def theta_color_correction(shadow_corrected_img,
mask,
save_ccimg_deg=False,
cd_save='./ccimg_deg_',
SNR_return=False,
save_SNR=False,
cd_SNR='./SNR_img'):
import cv2
from scipy.signal import find_peaks as fp
from functools import reduce
import numpy as np
import math
'''
Color Correction Using Degrees:
I_i^(shadow free) = ((r + 1)/(k_i*r + 1))I_i
r = L_d/L_e; L_d = intensity of non-shadow pixels
L_e = intensity of shadow pixels
k_i = t_i*cos(theta_i); t_i = 1 if object in sunshine region
and 0 if object in shadow region
(attenuation factor of direct light)
theta_i = angle between direct lighting
direction and the surface normal
I_i = pixel at i^th value
mask already gives us if object is in a sunshine region,
where t_i = 1, and if object is in a shadow region, where
t_i = 0.
Then iterate through varying values of theta to see how these
affect the outcome of the color correction to the image.
Need to find L_d and L_e, which seem to be global variables.
This doesn't seem like an especially smart move as there will
be many different intensities throughout the image, especially
with a very busy image.
Purpose of k_i is that if point is in a direct sunshine region, no angle between direct light and ambient light, then
that pixels value will not change as k_i will equal 1 due to being in a sunshine region and having a theta_i = 0. If
a pixel is in a shadow region, then pixel intensity should be brought up slightly to match sunshine region.
'''
##shadow_corrected_image should come from ycbcr_color_correction function
##mask is needed for SNR calculation
##save_ccimg_deg is optional to save all of the output pictures
##cd_save is where you want to save the 181 output pictures, as this function
##runs through all the angles from 0 to 180 to test on further correcting the colors of the image
##SNR_return will allow a return of the SNR calculated for each image
##Convert output image to HSI space to get purely intensity values of shadow to non-shadow pixels
result_HSI = cv2.cvtColor(shadow_corrected_img.copy(), cv2.COLOR_RGB2HSV)
##for SNR
sf_SNR = []
mask_rep = np.repeat(mask[:, :, np.newaxis], 3, axis=2)
##range of degrees from 0 to 180
deg_range = 181
for i in range(deg_range):
##intensity of shadow pixels
L_e_I = np.sum(np.multiply(result_HSI[:, :, 2],
mask.copy())) / (mask == 1).sum()
##intensity of non-shadow pixels
L_d_I = np.sum(
np.multiply(result_HSI[:, :, 2], np.logical_not(
mask.copy()))) / (mask == 0).sum()
##ratio of non-shadow pixel intensities to shadow pixel intensities
r_I = L_d_I / L_e_I
k_i = mask.copy() * math.cos(math.radians(i))
k_i[k_i == 0] = 1
result_shadow_free_I = np.uint8(
((r_I + 1) / (k_i * r_I + 1)) * result_HSI[:, :, 2].copy())
result_HSI[:, :, 2] = result_shadow_free_I.copy()
result_sf = cv2.cvtColor(result_HSI.copy(), cv2.COLOR_HSV2RGB)
result_sf = cv2.cvtColor(result_sf, cv2.COLOR_BGR2RGB)
##note that any pixels within the shadow region that are intrinsically
##zero will be discarded using this methodology, but this was the only way
##I figured to calculate SNR, specifically in the mask areas
result_sf_mask_area_mult = np.multiply(result_sf.copy(), mask_rep)
result_sf_mask_area_flat = result_sf_mask_area_mult.flat
result_sf_mask_area_fl8 = result_sf_mask_area_flat[
result_sf_mask_area_flat != 0]
sf_mean = np.mean(result_sf_mask_area_fl8.copy())
sf_std = np.std(result_sf_mask_area_fl8.copy())
ratio = sf_mean / sf_std
sf_SNR.append(ratio)
if save_ccimg_deg == True:
ccimg_deg = cd_save + str(i) + '.jpg'
cv2.imwrite(ccimg_deg, result_sf)
##finds mean of sf_SNR and divides by max value of sf_SNR to get unique threshold
thresh = (reduce(lambda x, y: x + y, sf_SNR) / 181) / (np.amax(sf_SNR))
SNR_peaks = fp(sf_SNR, height=np.amax(sf_SNR) - thresh, distance=10)
degrees = np.linspace(0, 180, 181)
plt.plot(degrees, sf_SNR)
plt.xlabel('degrees')
plt.ylabel('SNR')
plt.title('SNR vs degrees of ambient light to reflected light')
for i in range(SNR_peaks[0].shape[0]):
y_max = sf_SNR[SNR_peaks[0][i]]
x_max = SNR_peaks[0][i]
text_max = "x={:.3f}".format(x_max)
plt.annotate(text_max, xy=(x_max, y_max))
y_min = np.amin(sf_SNR)
x_min = sf_SNR.index(min(sf_SNR))
text_min = "x={:.3f}".format(x_min)
plt.annotate(text_min, xy=(x_min, y_min))
if save_SNR == True:
plt.savefig(cd_SNR + '.png', bbox_inches='tight')
if SNR_return == True:
return sf_SNR
r_v = np.array([(vel1)[zero_crossings], (vel1)[zero_crossings]])
Radial_Velocity = np.mean(r_v)
vel1 = vel1 - Radial_Velocity
vel2 = vel2 - Radial_Velocity
frequency, power = LombScargle(data[:, 0],
vel1).autopower(minimum_frequency=0.1,
maximum_frequency=2)
f_opt = frequency[np.where(power == np.max(power))]
def sine(x, V, f):
return V * np.sin(2 * np.pi * f * x)
indices1 = fp(np.abs(vel1), height=15)
v1 = np.mean(np.abs(vel1[indices1[0]]))
indices2 = fp(np.abs(vel2), height=12)
v2 = np.mean(np.abs(vel2[indices2[0]]))
p_opt = np.array([-v1, f_opt])
plt.subplot(2, 1, 1)
plt.title('Star 1')
plt.plot(data[:, 0], sine(data[:, 0], *p_opt), label='Best Fit', color='g')
plt.scatter(data[:, 0], vel1, s=0.1, color='darkorange')
p_opt = np.array([v2, f_opt])
plt.subplot(2, 1, 2)
plt.title('Star 2')
plt.plot(data[:, 0], sine(data[:, 0], *p_opt), label='Best Fit', color='g')
(s.h * s.c) / (l * s.k * T)) - 1))
return t
plt.plot(wvl, dat)
plt.show()
inp = input("Where to cut:\n")
if inp == 'max':
cut = np.argmax(dat)
else:
cut = wvl.tolist().index(int(inp))
cut = int(cut)
spec = wvl.tolist().index(3500)
val, var = cfit(B, l[cut:], dat[cut:], p0=[1000, 1])
peaks, _ = fp((B(l, val[0], val[1]) - dat), height=0, prominence=0.05)
wvlpk = (peaks * 5 + 1150)
specpk = []
for i in range(len(peaks)):
if wvlpk[i] > 6550 and wvlpk[i] < 6650:
specpk.append(wvlpk[i])
elif wvlpk[i] > 4750 and wvlpk[i] < 4850:
specpk.append(wvlpk[i])
elif wvlpk[i] > 4300 and wvlpk[i] < 4400:
specpk.append(wvlpk[i])
specpk = (np.array(specpk) + (-1150)) / 5
wvlpk = np.array(specpk).astype(int)
plt.plot((peaks * 5 + 1150), dat[peaks], "x")
plt.plot((wvlpk * 5 + 1150), dat[wvlpk], "o")
plt.plot(wvl, B(l, val[0], val[1]) - dat)
plt.plot(wvl, dat)
@author: Ryan
"""
import numpy as np
import matplotlib.pyplot as plt
from scipy.signal import find_peaks as fp
lc = np.genfromtxt("lightcurve_data.txt")
rv = np.genfromtxt("RV_data.txt")
jd = lc[:, 0] - lc[0, 0]
peaks = (lc[:, 1] - 1) * -1
# find peaks and subtract their x value to create folded rotation curve.
t = fp(peaks, height=0.007, width=5)
light = []
phase = []
for ind, i in enumerate(lc[:, 1]):
for ind2, j in enumerate(t[0]):
if j == ind:
for k in lc[ind - 10:ind + 10, 1]:
light.append(k)
for l in jd[ind - 10:ind + 10] - jd[ind]:
phase.append(l)
else:
continue
# plot the resulting light curve
for sat in range(CLASS[l].shape[1]):
nb_peaks1[K, sat] = P_HSENS1[l][K, sat, 0].shape[0]
nb_peaks2[K, sat] = P_HSENS2[l][K, sat, 0].shape[0]
nb_peaks3[K, sat] = P_HSENS3[l][K, sat, 0].shape[0]
nb_peaks4[K, sat] = P_HSENS4[l][K, sat, 0].shape[0]
NB_peaks1.append(nb_peaks1)
NB_peaks2.append(nb_peaks2)
NB_peaks3.append(nb_peaks3)
NB_peaks4.append(nb_peaks4)
Hue = np.arange(0, 2 * np.pi, 2 * np.pi / 24)
y = MM1[-1][29, -1, :]
ydet = np.concatenate((y[-2:], y))
fp(ydet,
height=(np.amax(y)) / 2,
distance=4,
prominence=(np.amax(y) / 8, y.max()))
R1 = F.peak_detection(y, show=False)
prop_null = np.zeros(len(CLASS))
prop_1 = np.zeros(len(CLASS))
prop_2 = np.zeros(len(CLASS))
prop_3 = np.zeros(len(CLASS))
prop_4 = np.zeros(len(CLASS))
for l in range(0, len(CLASS)):
a1 = np.amax(M_HSENS1[l], axis=-1) > t2
a2 = np.amax(M_HSENS2[l], axis=-1) > t2
a3 = np.amax(M_HSENS3[l], axis=-1) > t2
a4 = np.amax(M_HSENS4[l], axis=-1) > t2
ar = np.amax(M_ROT[l], axis=-1) > t2
