def dirFilter2D(mSize,nBands):
filts=[]
dirs=np.zeros((2,nBands),np.float)
theta=np.array(range(nBands))*math.pi/nBands
rho=np.ones(nBands)
X,Y=cv.polarToCart(rho,theta)
#X=np.cos(theta)
#Y=np.sin(theta)
dirs[0,:] =X.transpose()
dirs[1,:] =Y.transpose()
for k in np.array(range(nBands),np.float):
ang1 = (k-0.5)*math.pi/nBands;
ang2 = (k+ 0.5)*math.pi/nBands;
theta = np.array([ang1, ang2, ang1, ang2, ang1],float)
rho = np.array([1,1,-1,-1,1],float)*(mSize/2)
X,Y=cv.polarToCart(rho,theta)
#X=np.cos(theta)*rho
#Y=np.sin(theta)*rho
X=np.round(X+mSize/2)
Y=np.round(Y+mSize/2)
Mask=np.zeros((mSize,mSize),np.float)
polyVerticesTemp=np.array([X,Y],np.int32)
polyVertices=polyVerticesTemp.reshape(2,5)
polyVerticesNew=polyVertices.transpose()
Mask=cv.fillConvexPoly(Mask, polyVerticesNew, 1)
N=np.float(cv.countNonZero(Mask))
filts.append(Mask/N)
# correcting the orientation counter clockwise
temp = filts[1:]
temp.reverse()
filt0 = filts[0]
filts = [filt0] + temp
return filts, dirs
def _get_channel_sal_magn(self, channel):
if self.use_numpy_fft:
img_dft = np.fft.fft2(channel)
magnitude, angle = cv2.cartToPolar(np.real(img_dft),
np.imag(img_dft))
else:
img_dft = cv2.dft(np.float32(channel),
flags=cv2.DFT_COMPLEX_OUTPUT)
magnitude, angle = cv2.cartToPolar(img_dft[:, :, 0], img_dft[:, :,
1])
log_ampl = np.log10(magnitude.clip(min=1e-9))
log_ampl_blur = cv2.blur(log_ampl, (3, 3))
residual = np.exp(log_ampl - log_ampl_blur)
if self.use_numpy_fft:
real_part, imag_part = cv2.polarToCart(residual, angle)
img_combined = np.fft.ifft2(real_part + 1j * imag_part)
magnitude, _ = cv2.cartToPolar(np.real(img_combined),
np.imag(img_combined))
else:
img_dft[:, :, 0], img_dft[:, :,
1] = cv2.polarToCart(residual, angle)
img_combined = cv2.idft(img_dft)
magnitude, _ = cv2.cartToPolar(img_combined[:, :, 0],
img_combined[:, :, 1])
return magnitude
def main():
# Imagens
img1 = cv.imread("../imgs/puente.jpg", 0)
img2 = cv.imread("../imgs/ferrari-c.png", 0)
# Transformada
m1, f1 = tf_complexa(img1)
m2, f2 = tf_complexa(img2)
# Montando
x0, y0 = cv.polarToCart(m1, f2, angleInDegrees=False)
x1, y1 = cv.polarToCart(m2, f1, angleInDegrees=False)
im0 = cv.merge([x0, y0])
im1 = cv.merge([x1, y1])
# Inversa
inv0 = cv.idft(im0, cv.DFT_COMPLEX_OUTPUT)
inv1 = cv.idft(im1, cv.DFT_COMPLEX_OUTPUT)
# Combinar imagens para mostrar
r1 = cv.magnitude(inv0[:, :, 0], inv0[:, :, 1])
r2 = cv.magnitude(inv1[:, :, 0], inv1[:, :, 1])
# Normalizar
r1 = cv.normalize(r1, 0, 255, cv.NORM_MINMAX)
r2 = cv.normalize(r2, 0, 255, cv.NORM_MINMAX)
#escala logaritmica
#magn = cv.log(magn + 1)
#centralizar
#magn = np.fft.fftshift(magn, axes=None)
# Mostrar
plt.subplot(2, 2, 1)
plt.xticks([])
plt.yticks([])
plt.title("Imagem 1")
plt.imshow(img1, cmap="gray")
plt.subplot(2, 2, 2)
plt.xticks([])
plt.yticks([])
plt.title("Imagem 2")
plt.imshow(img2, cmap="gray")
plt.subplot(2, 2, 3)
plt.xticks([])
plt.yticks([])
plt.title("M1 + F2")
plt.imshow(r1, cmap="gray")
plt.subplot(2, 2, 4)
plt.xticks([])
plt.yticks([])
plt.title("M2 + F1")
plt.imshow(r2, cmap="gray")
plt.show()
def buildImgFromAmpMeanAndPse(amp, pse):
amp_mean = amp.copy().fill(amp.mean())
mixed = np.zeros((amp_mean.shape[0], amp_mean.shape[1], 2),
dtype=np.float64)
mixed[:mixed.shape[0], :mixed.shape[1],
0] = cv2.polarToCart(amp_mean, pse)[0]
mixed[:mixed.shape[0], :mixed.shape[1],
1] = cv2.polarToCart(amp_mean, pse)[1]
return cv2.idft(mixed, flags=cv2.DFT_REAL_OUTPUT | cv2.DFT_SCALE)
def mix(self, imageToBeMixed: 'ImageModel', magnitudeOrRealRatio: float,
phaesOrImaginaryRatio: float, mode: 'Modes') -> np.ndarray:
"""
a function that takes ImageModel object mag ratio, phase ration
"""
###
# implement this function
###
mix = np.zeros((self.imgByte.shape[0], self.imgByte.shape[1], 2),
'float64')
if mode.value == "testMagAndPhaseMode":
if (self.uniMag and imageToBeMixed.uniPh):
real, imaginary = cv.polarToCart(
self.uniMagnitude * magnitudeOrRealRatio +
imageToBeMixed.magnitude * (1 - magnitudeOrRealRatio),
self.phase * (1 - phaesOrImaginaryRatio) +
imageToBeMixed.uniPhase * phaesOrImaginaryRatio,
angleInDegrees=True)
elif (self.uniMag):
real, imaginary = cv.polarToCart(
self.uniMagnitude * magnitudeOrRealRatio +
imageToBeMixed.magnitude * (1 - magnitudeOrRealRatio),
self.phase * (1 - phaesOrImaginaryRatio) +
imageToBeMixed.phase * phaesOrImaginaryRatio,
angleInDegrees=True)
elif (imageToBeMixed.uniPh):
real, imaginary = cv.polarToCart(
self.magnitude * magnitudeOrRealRatio +
imageToBeMixed.magnitude * (1 - magnitudeOrRealRatio),
self.phase * (1 - phaesOrImaginaryRatio) +
imageToBeMixed.uniPhase * phaesOrImaginaryRatio,
angleInDegrees=True)
else:
real, imaginary = cv.polarToCart(
self.magnitude * magnitudeOrRealRatio +
imageToBeMixed.magnitude * (1 - magnitudeOrRealRatio),
self.phase * (1 - phaesOrImaginaryRatio) +
imageToBeMixed.phase * phaesOrImaginaryRatio,
angleInDegrees=True)
elif mode.value == "testRealAndImagMode":
real = self.real * magnitudeOrRealRatio + imageToBeMixed.real * (
1 - magnitudeOrRealRatio)
imaginary = self.imaginary * (
1 - phaesOrImaginaryRatio
) + imageToBeMixed.imaginary * phaesOrImaginaryRatio
mix[:, :, 0], mix[:, :, 1] = real, imaginary
invImg = cv.idft(mix, flags=cv.DFT_SCALE | cv.DFT_REAL_OUTPUT)
invImg *= 255.0 / np.max(invImg)
return invImg
def hybridImage(self):
#Gaussian method
trans1 = self.myGaussian(self.img1)
trans2 = self.img2 - self.myGaussian(self.img2)
result = trans2 + trans1
# result = result.astype(np.uint16)
cv.imwrite(OUTPUT_GAUSSIAN, result)
#FFT method
trans1 = self.myFFT(self.img1)
trans2 = self.myFFT(self.img2)
h = trans1.shape[0]
w = trans1.shape[1]
#中心点坐标
center_w = w // 2
center_h = h // 2
window = np.zeros((h, w))
for i in range(h):
for j in range(w):
r = ((i - center_h)**2 + (j - center_w)**2)**0.5
window[i][j] = self.get_cv(r, SIGMA_FFT)
#magnitude为幅值域,phase为相位域，幅值域与滤镜相乘，相位不变。
magnitude1 = cv.magnitude(trans1.real, trans1.imag)
magnitude2 = cv.magnitude(trans2.real, trans2.imag)
phase1 = cv.phase(trans1.real, trans1.imag)
phase2 = cv.phase(trans2.real, trans2.imag)
for i in range(3):
magnitude1[:, :, i] *= window
magnitude2[:, :, i] -= magnitude2[:, :, i] * window
#重新变为复数域
result1_real, result1_imag = cv.polarToCart(magnitude1, phase1)
result2_real, result2_imag = cv.polarToCart(magnitude2, phase2)
# print(magnitude)
# print(phase)
trans1.real = result1_real
trans1.imag = result1_imag
trans2.real = result2_real
trans2.imag = result2_imag
# print(trans1)
#反傅里叶变换
img_back = self.myIFFT(trans1 + trans2)
cv.imwrite(OUTPUT_FFT, img_back)
def polar(I, center, r, theta=(0, 360), rstep=1, thetastep=360.0 / (180 * 8)):
# 得到距离最小、最大范围。
minr, maxr = r
# 角度的最小范围
mintheta, maxtheta = theta
# 输出图像的高、宽
H = int((maxr - minr) / rstep + 1)
W = int((maxtheta - mintheta) / thetastep + 1)
O = 125 * np.ones((H, W), I.dtype)
cx, cy = center
# 极坐标变换
r = np.linspace(minr, maxr, H)
r = np.tile(r, (W, 1))
r = np.transpose(r) # 矩阵转置
theta = np.linspace(mintheta, maxtheta, W)
theta = np.tile(theta, (H, 1)) # 垂直方向重复H次，水平1次
x, y = cv2.polarToCart(r, theta, angleInDegrees=True)
# 最近邻插值
for i in range(H):
for j in range(W):
px = int(round(x[i][j]) + cx)
py = int(round(y[i][j]) + cy)
if ((px >= 0 and px <= W - 1) and (py >= 0 and py <= H - 1)):
O[i][j] = I[py][px]
else:
O[i][j] = 125 # 灰色
return O
def saliency_feature(img):
img_orig = img
img = cv2.resize(img, (64, 64))
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# h = cv2.getOptimalDFTSize(img.shape[0])
# w = cv2.getOptimalDFTSize(img.shape[1])
# print "Resizing (%d, %d) to (%d, %d)" % (img.shape[0], img.shape[1], h, w)
# h = (h - img.shape[0])/2.0
# w = (w - img.shape[1])/2.0
# img = cv2.copyMakeBorder(img, int(math.floor(h)), int(math.ceil(h)), int(math.floor(w)), int(math.ceil(w)), cv2.BORDER_CONSTANT, value=0)
dft = cv2.dft(np.float32(img), flags=cv2.DFT_COMPLEX_OUTPUT)
A, P = cv2.cartToPolar(dft[:,:,0], dft[:,:,1])
L = cv2.log(A)
h_n = (1./3**2)*np.ones((3,3))
R = L - cv2.filter2D(L, -1, h_n)
S = cv2.GaussianBlur(cv2.idft(np.dstack(cv2.polarToCart(cv2.exp(R), P)), flags=cv2.DFT_REAL_OUTPUT)**2, (0,0), 8)
S = cv2.resize(cv2.normalize(S, None, 0, 1, cv2.NORM_MINMAX), (img_orig.shape[1],img_orig.shape[0]))
# cv2.namedWindow('tmp1', cv2.WINDOW_NORMAL)
# cv2.imshow('tmp1', img_orig)
# cv2.namedWindow('tmp', cv2.WINDOW_NORMAL)
# cv2.imshow('tmp', S)
# cv2.waitKey()
return S
def re_transform(phases,amplitude):
x,y = cv.polarToCart(amplitude,phases)
back = cv.merge([x, y])
back_ishift = np.fft.ifftshift(back)
img_back = cv.idft(back_ishift)
img_back = cv.magnitude(img_back[:,:,0], img_back[:,:,1])
return img_back
def img_polar_transform(input_img,
center,
r_range,
theta_rangge=(0, 360),
r_step=0.5,
theta_step=360.0 / (90.)):
minr, maxr = r_range
mintheta, maxtheta = theta_rangge
H = int((maxr - minr) / r_step + 1)
W = int((maxtheta - mintheta) / theta_step + 1)
#print(H,W,theta_step)
output_img = np.zeros((H, W), input_img.dtype)
x_center, y_center = center
r = np.linspace(minr, maxr, H)
r = np.tile(r, (W, 1))
r = np.transpose(r)
theta = np.linspace(mintheta, maxtheta, W)
theta = np.tile(theta, (H, 1))
x, y = cv2.polarToCart(r, theta, angleInDegrees=True)
for i in range(H):
for j in range(W):
px = int(round(x[i, j]) + x_center)
py = int(round(y[i, j]) + y_center)
if ((px >= 0 and px <= 255) and (py >= 0 and py <= 255)):
#print(px,py)
output_img[i, j] = input_img[px, py]
return output_img
def update_pos_given_shift(self, single_marker):
print("Calculate single")
# Get the data of the new perceived marker
position_current_xy, distance_current, angle_surface_current = self.get_marker_xy(
single_marker)
# Calculate the difference in angle for coordination transform rotation
phi = single_marker.shift_angle_trans[0]
# Calculate the translation by calculating the coordinates according to the rotated origin
trans_pos_xy = cv2.polarToCart(distance_current, phi)
trans_pos_xy = (trans_pos_xy[0][0][0], trans_pos_xy[1][0][0])
# The translation vector is now the difference between the coordinates, relative to
# the base marker and the coordinates, relative to the new marker
trans_rel_base = single_marker.shift_angle_trans[1]
# Our position in the system of the base marker, given by the new marker, can now be calculated
# First the rotation is applied
new_x = position_current_xy[0] * np.cos(
phi) + position_current_xy[1] * np.sin(phi)
new_y = position_current_xy[0] * np.sin(
phi) + position_current_xy[1] * np.cos(phi)
# Then the translation is performed
resulting_x = new_x + trans_rel_base[0]
resulting_y = new_y + trans_rel_base[1]
# Write the resulting data into the marker
single_marker.pos_xy = (resulting_x, resulting_y)
def update_pos_shift_to_base(self, base_marker, marker_b):
# Get the data of the new perceived marker
position_current_xy, distance_current, angle_surface_current = self.get_marker_xy(
marker_b)
# Get the data of the base marker, perceived from the new position
base_xy, dist_base, ang_base = self.get_marker_xy(base_marker)
# Calculate the difference in angle for coordination transform rotation
phi = ang_base - angle_surface_current
# Calculate the translation by calculating the coordinates according to the rotated origin
trans_pos_xy = cv2.polarToCart(distance_current, phi)
trans_pos_xy = (trans_pos_xy[0][0][0], trans_pos_xy[1][0][0])
# The translation vector is now the difference between the coordinates, relative to
# the base marker and the coordinates, relative to the new marker
trans_rel_base = np.subtract(base_xy, trans_pos_xy)
# Our position in the system of the base marker, given by the new marker, can now be calculated
# First the rotation is applied
new_x = position_current_xy[0] * np.cos(
phi) + position_current_xy[1] * np.sin(phi)
new_y = position_current_xy[0] * np.sin(
phi) + position_current_xy[1] * np.cos(phi)
# Then the translation is performed
resulting_x = new_x + trans_rel_base[0]
resulting_y = new_y + trans_rel_base[1]
# Write the resulting data into the marker
marker_b.pos_xy = (resulting_x, resulting_y)
marker_b.shift_angle_trans = (phi, (trans_rel_base))
marker_b.aquired_at_distance = distance_current
def lens_distortion(pil_img, exp, scale):
img = np.array(pil_img)
img_c_2 = img[150:250, 100:200]
rows, cols = img_c_2.shape[:2]
mapy, mapx = np.indices((rows, cols), dtype=np.float32)
mapx = 2 * mapx / (cols - 1) - 1
mapy = 2 * mapy / (rows - 1) - 1
r, theta = cv2.cartToPolar(mapx, mapy)
r[r < scale] = r[r < scale]**exp
mapx, mapy = cv2.polarToCart(r, theta)
mapx = ((mapx + 1) * (cols - 1)) / 2
mapy = ((mapy + 1) * (rows - 1)) / 2
distorted = cv2.remap(img_c_2, mapx, mapy, cv2.INTER_LINEAR)
img[150:250, 100:200] = distorted
pil_img = Image.fromarray(img)
return pil_img
def compute_motion_compensation(motion_matrix, stab_mode,
trimmed_mean_percentage):
if stab_mode == 'mean':
u = motion_matrix[:, :, 0].mean()
v = motion_matrix[:, :, 1].mean()
elif stab_mode == 'trimmed_mean':
u = trim_mean(motion_matrix[:, :, 0],
trimmed_mean_percentage,
axis=None)
v = trim_mean(motion_matrix[:, :, 1],
trimmed_mean_percentage,
axis=None)
elif stab_mode == 'median':
u, v = np.median(motion_matrix[:, :, 0]), np.median(motion_matrix[:, :,
1])
elif stab_mode == 'mode':
us, vs = cv2.cartToPolar(motion_matrix[:, :, 0], motion_matrix[:, :,
1])
mu, mv = mode(us.ravel())[0], mode(vs.ravel())[0]
u, v = cv2.polarToCart(mu, mv)
u, v = u[0][0], v[0][0]
else:
raise NotImplemented(
"Choose one of implemented modes: mean, trimmed_mean, median")
return u, v
def arctan_unwarp(points, a, image_width, image_height):
"""
Transforms points to remove lens distortion according to an arctan model. Cannot produce the
identity function, but asymptotically approaches it as `a` goes toward 0
:param points:
ndarray of image space points
:param a:
real-valued scalar, tunable parameter.
:param image_width:
int, image width in pixels
:param image_height:
int, image image_height in pixels
:returns:
array of points after transformation
"""
x = points[..., 0]
y = points[..., 1]
# center
x_center = x - image_width / 2
y_center = y - image_height / 2
# convert to polar coords
r_polar, theta_polar = cv2.cartToPolar(x_center, y_center)
# # make lens correction
r_max = np.hypot(image_width, image_height) * 0.5
r_polar = r_max * np.tan(r_polar * np.arctan(a) / r_max) / a
# convert back to cartesian coords and un-center
x_cart, y_cart = cv2.polarToCart(r_polar, theta_polar)
x = np.squeeze(x_cart) + image_width / 2
y = np.squeeze(y_cart) + image_height / 2
# make array
points_adjusted = np.stack([x, y], axis=-1)
return points_adjusted
def polar(I,
center,
r,
theta=(0, 360),
r_step=1.0,
theta_step=360.0 / (180 * 8)):
"""
将圆形图进行极坐标转换，将截取的某段圆环拉升为矩形
"""
h, w = I.shape
#r的取值范围
min_r, max_r = r
#theta取值范围
min_theta, max_theta = theta
#输出图像的宽高
H = int((max_r - min_r) / r_step) + 1
W = int((max_theta - min_theta) / theta_step) + 1
out_pic = 125 * np.ones((H, W), I.dtype)
#极坐标转换
r = np.linspace(min_r, max_r, H)
r = np.tile(r, (W, 1))
r = np.transpose(r)
theta = np.linspace(min_theta, max_theta, W)
theta = np.tile(theta, (H, 1))
x, y = cv2.polarToCart(r, theta, angleInDegrees=1)
#最近邻插值
cx, cy = center
for i in xrange(H):
for j in xrange(W):
px = int(round(x[i][j]) + cx)
py = int(round(y[i][j]) + cy)
if ((px >= 0 and px <= w - 1)) and (py >= 0 and py <= h - 1):
out_pic[i][j] = I[px][py]
return out_pic
def img_polar_transform(input_img,
center=(IMAGE_SIZE / 2, IMAGE_SIZE / 2),
r_range=(0, int(IMAGE_SIZE / 2 * 1.414)),
theta_range=(0, 360),
r_step=1.0,
theta_step=360.0 / (360.0)):
minr, maxr = r_range
mintheta, maxtheta = theta_range
H = int((maxr - minr) / r_step + 1) #
W = int((maxtheta - mintheta) / theta_step + 1) #361
output_img = np.zeros((H, W), input_img.dtype)
x_center, y_center = center
r = np.linspace(minr, maxr, H)
r = np.tile(r, (W, 1))
r = np.transpose(r)
theta = np.linspace(mintheta, maxtheta, W)
theta = np.tile(theta, (H, 1))
x, y = cv2.polarToCart(r, theta, angleInDegrees=True)
for i in range(H):
for j in range(W):
px = int(round(x[i, j]) + x_center)
py = int(round(y[i, j]) + y_center)
#最后10
if ((px >= 0 and px <= IMAGE_SIZE - 1)
and (py >= 0 and py <= IMAGE_SIZE - 1)):
output_img[i, j] = input_img[px, py]
return output_img
def get_marker_xy(self, marker):
distance = aruco_data.get_distance(marker)
angle = aruco_data.get_angle_surface(marker)
#angle = aruco_data.get_angle_to_center(marker)
# Now we can get the dx and dy of movement
xy = cv2.polarToCart(distance, angle)
return tuple((xy[0][0][0], xy[1][0][0])), distance, angle
def inverse_fourier_transform(self):
plane = cv2.polarToCart(self.magnitude, self.phase)
fourier = cv2.merge(plane)
fourier_inv_shift = np.fft.ifftshift(fourier)
img_back = cv2.idft(fourier_inv_shift)
new_plane = cv2.split(img_back)
self.img_back = cv2.magnitude(new_plane[0], new_plane[1])
def polar(I,
center,
r,
theta=(0, 360),
rstep=1.0,
thetastep=360.0 / (180 * 8)):
#得到距离的最小最大范围
minr, maxr = r
#角度的最小范围
mintheta, maxtheta = theta
#输出图像的高，宽
H = int((maxr - minr) / rstep) + 1
W = int((maxtheta - mintheta) / thetastep) + 1
O = np.zeros((H, W), I.dtype)
#极坐标变换
r = np.linspace(minr, maxr, H)
r = np.tile(r, (W, 1))
r = np.transpose(r)
theta = np.linspace(mintheta, maxtheta, W)
theta = np.tile(theta, (H, 1))
x, y = cv2.polarToCart(r, theta, angleInDegrees=True)
#最近邻插值
for i in xrange(H):
for j in xrange(W):
px = int(round(x[i][j]) + cx)
py = int(round(y[i][j]) + cy)
if (px > w - 1 or py > h - 1):
O[i][j] = 125 #灰色
else:
O[i][j] = I[py][px]
return O
def test_Polar2Anix(self):
# lower_red(RGB) (208,79,113)
# upper (212,47,56)
r, angle = cv2.cartToPolar(1, 1)
print(r, angle)
x, y = cv2.polarToCart(r, angle)
print(x, y)
def saliency_feature(img):
img_orig = img
img = cv2.resize(img, (64, 64))
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# h = cv2.getOptimalDFTSize(img.shape[0])
# w = cv2.getOptimalDFTSize(img.shape[1])
# print "Resizing (%d, %d) to (%d, %d)" % (img.shape[0], img.shape[1], h, w)
# h = (h - img.shape[0])/2.0
# w = (w - img.shape[1])/2.0
# img = cv2.copyMakeBorder(img, int(math.floor(h)), int(math.ceil(h)), int(math.floor(w)), int(math.ceil(w)), cv2.BORDER_CONSTANT, value=0)
dft = cv2.dft(np.float32(img), flags=cv2.DFT_COMPLEX_OUTPUT)
A, P = cv2.cartToPolar(dft[:, :, 0], dft[:, :, 1])
L = cv2.log(A)
h_n = (1. / 3**2) * np.ones((3, 3))
R = L - cv2.filter2D(L, -1, h_n)
S = cv2.GaussianBlur(
cv2.idft(np.dstack(cv2.polarToCart(cv2.exp(R), P)),
flags=cv2.DFT_REAL_OUTPUT)**2, (0, 0), 8)
S = cv2.resize(cv2.normalize(S, None, 0, 1, cv2.NORM_MINMAX),
(img_orig.shape[1], img_orig.shape[0]))
# cv2.namedWindow('tmp1', cv2.WINDOW_NORMAL)
# cv2.imshow('tmp1', img_orig)
# cv2.namedWindow('tmp', cv2.WINDOW_NORMAL)
# cv2.imshow('tmp', S)
# cv2.waitKey()
return S
def get_center_circle_points(own_xys):
goalxys = []
goal_offset = -(np.pi / 8.0)
circle_radius = radius_arena - distance_from_boundary_of_circle # meter
# calculate positions on a circle near the center
for pos in own_xys:
own_x = pos[0]
own_y = pos[1]
# Get a point on the circle
distance_goalpoint = circle_radius
# Now take the distance to the goalpoint in combination with the angle to the
# goalpoint to make out the intersection of the straight line in between the
# car and that circle
angle_center = np.arctan2(own_y, own_x)
# Change the angle to be in front of the vehicle like the carrot on a stick
angle_goalpoint = angle_center + goal_offset
# Next calculate that new point
goal_xy = cv2.polarToCart(distance_goalpoint, angle_goalpoint)
goal_xy = [goal_xy[0][0][0], goal_xy[1][0][0]]
# If the distance of that new goalpoint is too far from our position
goalxys.append(goal_xy)
return goalxys
def bin_sum(h_var, ventana):
vector = np.arange(360)
vector.resize(1, 360)
X, Y = cv2.polarToCart(h_var, vector * (np.pi / 180))
X_nueva = np.zeros((1, 360))
Y_nueva = np.zeros((1, 360))
for i in range(0, 360):
if ((i + ventana - 1) > 360):
topeX = (i + ventana - 1) - 360
topeY = (i + ventana - 1) - 360
X1 = X[0, i:360].reshape(i - 360, 1)
X2 = X[0, 0:topeX].reshape(topeY, 1)
if X1.shape[0] == 0:
XX = X2
elif X2.shape[0] == 0:
XX = X1
else:
XX = np.append(X1, X2)
Y1 = Y[0, i:360].reshape(i - 360, 1)
Y2 = Y[0, 0:topeY].reshape(topeY, 1)
if Y1.shape[0] == 0:
YY = Y2
elif Y2.shape[0] == 0:
YY = Y1
else:
YY = np.append(Y1, Y2)
else:
XX = X[0, i:i + ventana - 1]
YY = Y[0, i:i + ventana - 1]
pond = np.linspace(0, 1, ventana - 1)
X_nueva[0, i] = np.sum(XX * pond) / ventana
Y_nueva[0, i] = np.sum(YY * pond) / ventana
rho, theta = cv2.cartToPolar(X_nueva, Y_nueva)
return rho, theta
def onChanged(x):
exp = x / 10
# 매핑 배열 생성 ---②
mapy, mapx = np.indices((rows, cols), dtype=np.float32)
# 좌상단 기준좌표에서 -1~1로 정규화된 중심점 기준 좌표로 변경 ---③
mapx = (2 * mapx - cols) / cols
mapy = (2 * mapy - rows) / rows
# 직교좌표를 극 좌표로 변환 ---④
r, theta = cv2.cartToPolar(mapx, mapy)
# 왜곡 영역만 중심확대/축소 지수 적용 ---⑤
r[r < scale] = r[r < scale]**exp
# 극 좌표를 직교좌표로 변환 ---⑥
mapx, mapy = cv2.polarToCart(r, theta)
# 중심점 기준에서 좌상단 기준으로 변경 ---⑦
mapx = ((mapx + 1) * cols) / 2
mapy = ((mapy + 1) * rows) / 2
# 재매핑 변환
distorted = cv2.remap(img, mapx, mapy, cv2.INTER_LINEAR)
cv2.imshow('lens', distorted)
def car_to_field(self, rvec, tvec):
"""Transform the car's location and direction to field coordinates, shift to field plane.
:param rvec: Rotation vector (as output from aruco.estimatePoseBoard)
:param tvec: Translation vector (as output from aruco.estimatePoseBoard)
:return: loc, dir - Location in field coordinates, direction vector (normalized) in field coords.
"""
cam_mat = cam_params.logitech_matrix
cam_dist = cam_params.logitech_dist_coeffs
mat = np.frombuffer(self.master.perspective_matrix.get_obj()).reshape(
(3, 3))
if mat.any(
): # If the perspective matrix has already been updated by StaticProcessor
obj_points = np.array([[0, 0, 0], [0, 1, 0]], np.float32)
pic_points, _ = cv2.projectPoints(obj_points, rvec, tvec, cam_mat,
cam_dist)
cx, cy = self.master.center.coords()
lens, angles = cv2.cartToPolar(cx - pic_points[:, 0, 0],
cy - pic_points[:, 0, 1])
lens *= (1 - 85. / H_CAM_MM)
pic_points = np.array(cv2.polarToCart(lens, angles)).T
pic_points = pic_points[0, :, None] # Cause OpenCV is ...
pic_points = np.array((cx, cy)) - pic_points
field_points = cv2.perspectiveTransform(pic_points, mat)
loc = field_points[0, 0]
dir = field_points[1, 0] - field_points[0, 0]
cv2.normalize(dir, dir, 70 * pixres, norm_type=2)
return loc, dir
else:
return None, None
def get_fov_triangle(self, xy, heading, fov_angle, distance):
'''
returns the field of view as triangle, based on a sequence of
coordinates. It will always regard the last coordinates in that list
'''
limit_left = heading - np.deg2rad(fov_angle / 2.0)
limit_right = heading + np.deg2rad(fov_angle / 2.0)
a = xy # Point a is the latest known position
b = cv2.polarToCart(distance, limit_left) # Point b is distance in front and half of fov to the left
c = cv2.polarToCart(distance, limit_right) # Point c is distance in front and half of fov to the right
# Needed extraction of the values from the polarToCart method
# and adding of a as the origin
b = b[0][0][0] + a[0], b[1][0][0] + a[1]
c = c[0][0][0] + a[0], c[1][0][0] + a[1]
return Triangle(to_point(a), to_point(b), to_point(c))
def project_pos(xy, current_heading, distance=2):
'''
Take the position as xy and the current heading and returns future
xy in a distance of distance
'''
x, y = cv2.polarToCart(distance, current_heading)
return [x[0][0] + xy[0], y[0][0] + xy[1]]
def _get_channel_sal_magn(self, channel):
"""Returns the log-magnitude of the Fourier spectrum
This method calculates the log-magnitude of the Fourier spectrum
of a single-channel image. This image could be a regular grayscale
image, or a single color channel of an RGB image.
:param channel: single-channel input image
:returns: log-magnitude of Fourier spectrum
"""
# do FFT and get log-spectrum
if self.use_numpy_fft:
img_dft = np.fft.fft2(channel)
magnitude, angle = cv2.cartToPolar(np.real(img_dft),
np.imag(img_dft))
else:
img_dft = cv2.dft(np.float32(channel),
flags=cv2.DFT_COMPLEX_OUTPUT)
magnitude, angle = cv2.cartToPolar(img_dft[:, :, 0],
img_dft[:, :, 1])
# get log amplitude
log_ampl = np.log10(magnitude.clip(min=1e-9))
# blur log amplitude with avg filter
log_ampl_blur = cv2.blur(log_ampl, (3, 3))
# residual
residual = np.exp(log_ampl - log_ampl_blur)
# back to cartesian frequency domain
if self.use_numpy_fft:
real_part, imag_part = cv2.polarToCart(residual, angle)
img_combined = np.fft.ifft2(real_part + 1j*imag_part)
magnitude, _ = cv2.cartToPolar(np.real(img_combined),
np.imag(img_combined))
else:
img_dft[:, :, 0], img_dft[:, :, 1] = cv2.polarToCart(residual,
angle)
img_combined = cv2.idft(img_dft)
magnitude, _ = cv2.cartToPolar(img_combined[:, :, 0],
img_combined[:, :, 1])
return magnitude
def remove_periodic_interference(self, image):
height, width = image.shape
# min and max values of image
min_v = np.amin(image)
max_v = np.amax(image)
mean_v = int(np.mean(image))
# adding padding to image
padded_height = math.ceil(math.log2(height))
padded_height = int(math.pow(2, padded_height))
padded_width = math.ceil(math.log2(width))
padded_width = int(math.pow(2, padded_width))
padded_image = np.full((padded_height, padded_width), mean_v, dtype=np.uint8)
padded_image[0:height, 0:width] = image
# convert image to float and save as complex output
dft = cv2.dft(np.float32(padded_image), flags=cv2.DFT_COMPLEX_OUTPUT)
# shift of origin from upper left corner to center of image
dft_shifted = np.fft.fftshift(dft)
# magnitude and phase images
mag, phase = cv2.cartToPolar(dft_shifted[:, :, 0], dft_shifted[:, :, 1])
# extract spectrum
spectrum = np.log(mag) / 20
min, max = np.amin(spectrum, (0, 1)), np.amax(spectrum, (0, 1))
# threshold spectrum to find bright spots
thresh_spec = (255 * spectrum).astype(np.uint8)
thresh_spec = cv2.threshold(thresh_spec, 155, 255, cv2.THRESH_BINARY)[1]
# cover center rows of thresh with black
yc = padded_height // 2
cv2.line(thresh_spec, (0, yc), (padded_width - 1, yc), 0, 5)
# get y coordinates bright spots
bright_spots = np.column_stack(np.nonzero(thresh_spec))
# mask
mask = thresh_spec.copy()
for b in bright_spots:
y = b[0]
cv2.line(mask, (0, y), (padded_width - 1, y), 255, 5)
# apply mask to magnitude
mag[mask != 0] = 0
# convert new magnitude and old phase into cartesian real and imaginary components
real, imag = cv2.polarToCart(mag, phase)
# combine cart comp into one compl image
back = cv2.merge([real, imag])
# shift origin from center to upper left corner
back_ishift = np.fft.ifftshift(back)
# do idft saving as complex output
img_back = cv2.idft(back_ishift)
# combine complex components into original image again
img_back = cv2.magnitude(img_back[:, :, 0], img_back[:, :, 1])
# crop to original size
img_back = img_back[0:height, 0:width]
# re-normalize to 8-bits in range of original
min, max = np.amin(img_back, (0, 1)), np.amax(img_back, (0, 1))
notched = cv2.normalize(img_back, None, alpha=min_v, beta=max_v,
norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8U)
return notched
def computeSaliencyImg(img):
"""
This function computes the saliency map of a given image per the algorithm
put forth in the work of Xiaodi Hou et. al.
params:
img (int8 image): the image to compute the saliency map for.
returns:
saliencyMap (int8 image): the computed saliency map
"""
# compute dft and shift
fImage = cv2.dft(np.float32(img), flags=cv2.DFT_COMPLEX_OUTPUT)
fImageShifted = np.fft.fftshift(fImage)
# extract real and imaginary image components in polar form
real, imaginary = cv2.cartToPolar(fImageShifted[:, :, 0],
fImageShifted[:, :, 1])
# compute log of real image
logImage = np.log(real)
# compute blurred version of real image
blurKernel = np.ones((7, 7), np.float32) / (7 * 7)
logImageBlured = cv2.filter2D(logImage, -1, blurKernel)
# compute spectral residue image on log scale
spectralResidualImageLog = logImage - logImageBlured
# undo log scaling to real image component matches non-log scaling
# of phase/imaginary image component
spectralResidualImage = np.exp(spectralResidualImageLog)
# move back to cartesian complex form
fImageShifted[:, :, 0], fImageShifted[:, :, 1] = cv2.polarToCart(
spectralResidualImage, imaginary)
# compute neccessary shift for invese DFT
fInvShift = np.fft.ifftshift(fImageShifted)
# perform inverse DFT
saliencyMapComplex = cv2.idft(fInvShift)**2
# find real part of saliency map from obtained DFT results
saliencyMap = cv2.magnitude(saliencyMapComplex[:, :, 0],
saliencyMapComplex[:, :, 1])
# normalize around the maximum value (result between 0-1)
maxVal = np.max(saliencyMap)
saliencyMap = saliencyMap / maxVal
# scale image up to 8-bit range (0-255)
saliencyMap = saliencyMap * 255
return saliencyMap
def __GetChannelSalMagn(self, channel):
# do FFT and get log-spectrum
if self.useNumpyFFT:
imgDFT = np.fft.fft2(channel)
magnitude,angle = cv3.cartToPolar(np.real(imgDFT), np.imag(imgDFT))
else:
imgDFT = cv3.dft(np.float32(channel), flags=cv3.DFT_COMPLEX_OUTPUT)
magnitude,angle = cv3.cartToPolar(imgDFT[:,:,0], imgDFT[:,:,1])
# get log amplitude
logAmpl = np.log10(magnitude.clip(min=1e-9))
# blur log amplitude with avg filter
logAmplBlur = cv3.blur(logAmpl, (3,3))
# residual
residual = np.exp(logAmpl - logAmplBlur)
# back to cartesian frequency domain
if self.useNumpyFFT:
realPart,imagPart = cv3.polarToCart(residual, angle)
imgCombined = np.fft.ifft2(realPart + 1j*imagPart)
magnitude,_ = cv3.cartToPolar(np.real(imgCombined), np.imag(imgCombined))
else:
imgDFT[:,:,0],imgDFT[:,:,1] = cv3.polarToCart(residual, angle)
imgCombined = cv3.idft(imgDFT)
magnitude,_ = cv3.cartToPolar(imgCombined[:,:,0], imgCombined[:,:,1])
# magnitude = magnitude - np.mean(magnitude)
# if self.gaussKernel is not None:
# magnitude = cv3.GaussianBlur(np.float32(magnitude), self.gaussKernel, sigmaX=8, sigmaY=0)
# magnitude = magnitude**2
# magnitude = np.float32(magnitude)/np.max(magnitude)
return magnitude
def _make_mapping_matrix(self): paintable_area = 0.95 * (self._window_size / 2.0 - self._annulus) angles = np.zeros((Screen.screen_vane_count, Screen.screen_max_magnitude)) magnitudes = np.zeros((Screen.screen_vane_count, Screen.screen_max_magnitude)) for angle in xrange(Screen.screen_vane_count): for mag in xrange(Screen.screen_max_magnitude): render_angle = (angle+0.5) / float(Screen.screen_vane_count) * 2.0 * 3.14159 render_mag = self._annulus + (Screen.screen_max_magnitude - mag) / \ float(Screen.screen_max_magnitude) * paintable_area angles[angle, mag] = render_angle magnitudes[angle, mag] = render_mag cols, rows = cv2.polarToCart(magnitudes, angles) cols = np.round((cols + self._window_size / 2)).astype(np.int32) rows = np.round((rows + self._window_size / 2)).astype(np.int32) return cols, rows
def polar_transformation(image, origin, size, distance_range, angle_range=(0, 2*np.pi)):
h, w = image.shape[0:2]
distance = np.linspace(distance_range[0], distance_range[1], size[0])
angle = np.linspace(angle_range[0],angle_range[1],size[1])
angle_arr, distance_arr = np.meshgrid(angle, distance)
xv, yv = cv2.polarToCart(distance_arr, angle_arr)
xv += origin[1]
yv += origin[0]
yv = np.clip(yv, 0, h - 1)
xv = np.clip(xv, 0, w - 1)
if len(image.shape) == 3:
polar_image = image[yv.astype('int32'), xv.astype('int32'), :]
else:
polar_image = image[yv.astype('int32'), xv.astype('int32')]
return polar_image
print dft1.shape
dft_shift1 = np.fft.fftshift(dft1)
dftmag1, dftphase1 = cv2.cartToPolar(dft_shift1[:,:,0], dft_shift1[:,:,1])
invdftmag = np.ones(dftmag1.shape)
invdftphase = np.zeros(dftphase1.shape)
for i in range(-240,241):
for j in range(-240,241):
if np.exp(-0.0025*(i*i+j*j)**(5/6)) >= 0.01:
invdftmag[i,j] = np.exp(0.0025*(i*i+j*j)**(5/6))
print invdftmag
dftmag2 = invdftmag*dftmag1
dftphase2 = dftphase1 - invdftphase
dft_shift1[:,:,0], dft_shift1[:,:,1] = cv2.polarToCart(dftmag2, dftphase2)
f_ishift = np.fft.ifftshift(dft_shift1)
img2 = cv2.idft(f_ishift)
img2 = cv2.magnitude(img2[:,:,0],img2[:,:,1])
print img1
print img2/np.max(img2)*255
plt.subplot(121),plt.imshow(img1, cmap = 'gray')
plt.title('Input'), plt.xticks([]), plt.yticks([])
plt.subplot(122),plt.imshow(np.uint8(img2/np.max(img2)*255), cmap = 'gray')
plt.title('Restored'), plt.xticks([]), plt.yticks([])
plt.show()
def animate(i):
global opts, framegrabber, frames, params, flowStartFrame
global roi, flowMask, lmask, rmask, tmask, bmask
global flowVals, times, history
global logfile
global codes
update = [imdisp, foedisp, q_foe, b_latdiv, b_verdiv, b_ttc]
t1 = time.time()
# ------------------------------------------------------------
# Compute optical flow
# ------------------------------------------------------------
# grab the current frame, update indices
clrframe = framegrabber.next()
currFrame = cv2.cvtColor(clrframe, cv2.COLOR_BGR2GRAY)
framenum = i + flowStartFrame
times[i] = framenum
prvs = sum(frames) / float(opts.frameavg)
nxt = (sum(frames[1:]) + currFrame) / float(opts.frameavg)
flow = cv2.calcOpticalFlowFarneback(prvs[startY:stopY,startX:stopX]
, nxt[startY:stopY,startX:stopX]
, **params)
mag, angle = cv2.cartToPolar(flow[...,0], flow[...,1])
# ------------------------------------------------------------
# Remove outlier flow
# ------------------------------------------------------------
if opts.nofilt:
thresh_mask = flowMask
else:
# clean up flow estimates, remove outliers
# thresh_mask = threshold_local(mag, shape=(20,20), llim=0, ulim=0.96)
# global_mask = threshold_global(mag, llim=0.00, ulim=0.99)[0]
global_mask = np.ones_like(flowMask)
lthresh = 1e-3
thresh_mask = (mag > lthresh) & global_mask
flow[~thresh_mask] = 0
mag[~thresh_mask] = 0
# ------------------------------------------------------------
# estimate the location of the FoE
# ------------------------------------------------------------
# S = generic2dFilter(angle, (foeW, foeW), matchWin, step=dt, padded=True)
# participants = generic2dFilter(thresh_mask, (foeW, foeW), np.sum
# , padded=True, step=dt)
# S /= participants
# foe_y_subsearch, foe_x_subsearch = np.unravel_index(np.argmin(S), S.shape)
# foe_y, foe_x = startY + foeW//2 + foe_y_subsearch*dt , startX + foeW//2 + foe_x_subsearch*dt
# foe_x, foe_y = FindFoE(flow[...,0][foeW//2:-foeW//2],flow[...,1][foeW//2:-foeW//2])
# print
# print foe_x, foe_y
# foe_x, foe_y = startX + maskW//2, startY + maskW//2
foe_x, foe_y = 183, 80
p0, p1 = (foe_x-foeW//2, foe_y-foeW//2), (foe_x+foeW//2, foe_y+foeW//2)
# confidence= participants[foe_y_subsearch, foe_x_subsearch] / (foeW**2)
confidence=0
divTemplates = generate2dTemplates(p0, p1, thresh_mask.shape, thresh_mask)
foe_tmask, foe_bmask, foe_lmask, foe_rmask = divTemplates
foeSlice_y, foeSlice_x = slice(p0[1],p1[1]+1), slice(p0[0],p1[0]+1)
# ------------------------------------------------------------
# estimate divergence parameters and ttc for this frame
# ------------------------------------------------------------
xDiv = (np.sum(flow[rmask,0]) - np.sum(flow[lmask,0])) \
/(np.sum((lmask|rmask) & thresh_mask) + EPS)
yDiv = (np.sum(flow[tmask,1]) - np.sum(flow[bmask,1])) \
/(np.sum((tmask|bmask) & thresh_mask) + EPS)
xDiv_foe = (np.sum(flow[foe_rmask,0]) - np.sum(flow[foe_lmask,0])) \
/(np.sum(foe_lmask|foe_rmask) + EPS)
yDiv_foe = (np.sum(flow[foe_tmask, 1]) - np.sum(flow[foe_bmask, 1])) \
/(np.sum(foe_tmask|foe_bmask) + EPS)
ttc = 2/(xDiv + yDiv + EPS)
history[:, :-1] = history[:, 1:]; history[:, -1] = (xDiv,yDiv,ttc)
# ------------------------------------------------------------
# use estimation history to estimate new values
# ------------------------------------------------------------
if i > history.shape[1]:
flowVals[:-1, i] = np.sum(history[:-1]*w_forget, axis=1)/sum(w_forget)
# flowVals[-1, i] = np.median(history[-1,-3:])
# m, y0, _, _, std = stats.linregress(np.arange(5)
# , flowVals[-1,i-4:i+1]*w_forget[:5]/sum(w_forget[:5]))
# flowVals[2, i] = m*times[i] + y0
# flowVals[2, i] = np.median(history[-1, -3:])
flowVals[2, i] = stats.trim_mean(history[-1, -5:],0.4)
else:
flowVals[:, i] = (xDiv,yDiv,ttc)
# ------------------------------------------------------------
# write out results
# ------------------------------------------------------------
t2 = time.time()
out = (framenum, xDiv, yDiv, ttc, 100.*confidence, t2-t1)
if opts.log:
print >> logfile, ','.join(map(str,out))
if not opts.quiet:
sys.stdout.write("\r%4d %+6.2f %+6.2f %6.2f %6.2f %6.2f" % out)
sys.stdout.flush()
# ------------------------------------------------------------
# update figure
# ------------------------------------------------------------
b_latdiv.set_height(flowVals[0,i])
b_verdiv.set_height(flowVals[1,i])
b_ttc.set_height(flowVals[2,i])
foedisp.set_data(clrframe[foeSlice_y, foeSlice_x, ::-1].copy())
# clrframe[mag <= lthresh, :] = codes.colors[0][::-1]
# clrframe[~global_mask, :] = codes.colors[-1][::-1]
cv2.rectangle(clrframe, p0, p1, color=(0,255,0))
cv2.rectangle(clrframe, p0, (foe_x, foe_y+foeW//2), color=(255,0,0))
if opts.vis == "color_overlay":
cf.colorFlow(flow, clrframe[...,::-1]
, slice(startX,stopX), slice(startY,stopY), thresh_mask)
dispim = clrframe[..., ::-1]
elif opts.vis == "color":
dispim = cf.flowToColor(flow)
elif opts.vis == "quiver":
update.append(q_img) # add this object to those that are to be updated
q_img.set_UVC(flow[flow_strides, flow_strides, 0]
, flow[flow_strides, flow_strides, 1]
, (mag[flow_strides, flow_strides] \
*255/(np.max(mag)-np.min(mag)+EPS)))
dispim = clrframe[..., ::-1]
imdisp.set_data(dispim)
unitmag = 2*np.ones(foedisp.get_size())
foeKern = generateFoEkernel(foeW)
foeKern[foeW//2, foeW//2] = angle[foe_y,foe_x]
sim = (foeKern-angle[foeSlice_y,foeSlice_x])**2
X, Y = cv2.polarToCart(unitmag, angle[foeSlice_y,foeSlice_x].astype(np.float))
q_foe.set_UVC(X[1:-1:2, 1:-1:2], Y[1:-1:2, 1:-1:2]
, (sim[1:-1:2, 1:-1:2] \
*255/(np.max(sim)-np.min(sim)+EPS)))
# shift the frame buffer
frames[:-1] = frames[1:]; frames[-1] = currFrame
return update
def __init__(
self,
numNeurons=1000,
timeNow=0.0,
neuronTypes=[Neuron],
bounds=default_bounds,
neuronLocations=None,
distribution=default_distribution,
**kwargs
):
self.id = Population.id_ctr
Population.id_ctr += 1
self.numNeurons = numNeurons
self.timeNow = timeNow
self.neuronTypes = neuronTypes
self.bounds = bounds
self.center = (self.bounds[0] + self.bounds[1]) / 2
self.distribution = distribution
self.isConnected = False
self.plotColor = population_plot_colors[self.id % len(population_plot_colors)] # [graph]
self.inhibitoryConnectionColor = inhibitory_connection_color # [graph]
self.logger = logging.getLogger(
self.__class__.__name__
) # we could use "{}.{}".format(self.__class__.__name__, self.id) instead, but that'll create separate loggers for each Population
self.logger.info("Creating {}".format(self))
self.logger.debug("Bounds: x: {}, y: {}, z: {}".format(self.bounds[:, 0], self.bounds[:, 1], self.bounds[:, 2]))
# * Designate neuron locations
if neuronLocations is not None:
self.neuronLocations = neuronLocations
else:
self.neuronLocations = []
if isinstance(self.distribution, MultivariateUniform):
# NOTE self.distribution has to be a 3-channel MultivariateUniform, even if the third channel is a constant (low=high)
self.neuronLocations = np.column_stack(
[
np.random.uniform(self.distribution.lows[0], self.distribution.highs[0], self.numNeurons),
np.random.uniform(self.distribution.lows[1], self.distribution.highs[1], self.numNeurons),
np.random.uniform(self.distribution.lows[2], self.distribution.highs[2], self.numNeurons),
]
)
# self.logger.debug("MultivariateUniform array shape: {}".format(self.neuronLocations.shape))
elif isinstance(self.distribution, MultivariateNormal):
# self.logger.debug("Distribution: mu: {}, cov: {}".format(self.distribution.mu, self.distribution.cov)) # ugly
self.neuronLocations = np.random.multivariate_normal(
self.distribution.mu, self.distribution.cov, self.numNeurons
)
elif isinstance(self.distribution, SymmetricNormal):
thetas = np.random.uniform(pi, -pi, self.numNeurons) # symmetric in any direction around Z axis
rads = np.random.normal(
self.distribution.mu, self.distribution.sigma, self.numNeurons
) # varies radially
xLocs, yLocs = cv2.polarToCart(rads, thetas)
zLocs = np.repeat(np.float32([self.distribution.center[2]]), self.numNeurons).reshape(
(self.numNeurons, 1)
) # constant z, repeated as a column vector
# self.logger.debug("SymmetricNormal array shapes:- x: {}, y: {}, z: {}".format(xLocs.shape, yLocs.shape, zLocs.shape))
self.neuronLocations = np.column_stack(
[self.distribution.center[0] + xLocs, self.distribution.center[1] + yLocs, zLocs]
) # build Nx3 numpy array
elif isinstance(self.distribution, SymmetricLogNormal):
thetas = np.random.uniform(pi, -pi, self.numNeurons) # symmetric in any direction around Z axis
rads = np.random.lognormal(
self.distribution.mu, self.distribution.sigma, self.numNeurons
) # varies radially
xLocs, yLocs = cv2.polarToCart(rads, thetas)
zLocs = np.repeat(np.float32([self.distribution.center[2]]), self.numNeurons).reshape(
(self.numNeurons, 1)
) # constant z, repeated as a column vector
# self.logger.debug("SymmetricLogNormal array shapes:- x: {}, y: {}, z: {}".format(xLocs.shape, yLocs.shape, zLocs.shape))
self.neuronLocations = np.column_stack(
[self.distribution.center[0] + xLocs, self.distribution.center[1] + yLocs, zLocs]
) # build Nx3 numpy array
else:
raise ValueError("Unknown distribution type: {}".format(type(self.distribution)))
# TODO Include (non-central) F distribution (suitable for rods)
# Clip (clamp) neuron locations that are outside bounds
np.clip(self.neuronLocations[:, 0], self.bounds[0, 0], self.bounds[1, 0], out=self.neuronLocations[:, 0])
np.clip(self.neuronLocations[:, 1], self.bounds[0, 1], self.bounds[1, 1], out=self.neuronLocations[:, 1])
# print "Out-of-bounds neuron locations:", [loc for loc in self.neuronLocations if not ((self.bounds[0, 0] <= loc[0] <= self.bounds[1, 0]) and (self.bounds[0, 1] <= loc[1] <= self.bounds[1, 1]))] # [debug]
# print "Neuron locations:\n", self.neuronLocations # [debug]
# * Create neurons
self.neurons = self.numNeurons * [None]
self.neuronPlotColors = self.numNeurons * [None]
for i in xrange(self.numNeurons):
self.neurons[i] = random.choice(self.neuronTypes)(self.neuronLocations[i], self.timeNow, **kwargs)
self.neuronPlotColors[i] = self.neurons[i].plotColor
# * Build spatial index using quadtree (assuming neurons are roughly in a layer)
boundingRect = (self.bounds[0, 0], self.bounds[0, 1], self.bounds[1, 0], self.bounds[1, 1])
self.qtree = QuadTree(self.neurons, depth=int(log(self.numNeurons, 2)), bounding_rect=boundingRect)
image_to_analyze = cv2.imread('/home/frederik/pcl/images/perspective_result.jpeg', 0)
src_f = np.array(image_to_analyze, dtype=np.float32)
src_f /= 255.
dftl = cv2.dft(src_f,flags = cv2.DFT_COMPLEX_OUTPUT)
da, fa = cv2.split(dftl)
ma, pa = cv2.cartToPolar(da, fa)
ma = np.fft.fftshift(ma)
result = m * ma
ma = np.fft.fftshift(result)
d2, f2 = cv2.polarToCart(ma, pa)
test = np.zeros( (d.shape[0] , d.shape[1] , 2), dtype=np.float)
test[:,:,0] = d2
test[:,:,1] = f2
img_f = cv2.dft(test,flags = cv2.DFT_INVERSE + cv2.DFT_SCALE + cv2.DFT_REAL_OUTPUT)
cv2.imshow("Filter ", img_f )
#m = m * im
#cv2.imshow("Magnitude_after", np.uint8( 20*np.log(m) ) )
#cv2.imshow("Filter", np.uint8(im*255))
def dirFilter2D(mSize,nBands):
filts=[]
dirs=np.zeros((2,nBands),np.float)
theta=np.array(range(nBands))*math.pi/nBands
rho=np.ones(nBands)
X,Y=cv.polarToCart(rho,theta)
#X=np.cos(theta)
#Y=np.sin(theta)
dirs[0,:] =X.transpose()
dirs[1,:] =Y.transpose()
for k in np.array(range(nBands),np.float):
ang1 = (k-0.5)*math.pi/nBands;
ang2 = (k+ 0.5)*math.pi/nBands;
theta = np.array([ang1, ang2, ang1, ang2, ang1],float)
if flag==0:
#triangular section generation
Ang1=k*math.pi/nBands;
Ang2=(k+1)*math.pi/nBands;
Theta=np.array([Ang1,Ang2],float)
Rho1=np.array([1,1],float)*math.floor(mSize/2)
# xCor,yCor=cv.polarToCart(Rho,Theta)
x=Rho1*np.cos(Theta)+math.ceil(mSize/2)
y=Rho1*np.sin(Theta)+math.ceil(mSize/2)
Mask1=np.zeros((mSize,mSize),np.float)
polyVerticesTemp=np.array(np.round([[x[0],y[0]],[x[1],y[1]],[mSize/2,mSize/2]]),np.int32)
# polyVertices=polyVerticesTemp.reshape(2,3)
# polyVerticesNew=polyVertices.transpose()
Mask1=cv.fillConvexPoly(Mask1, polyVerticesTemp, 1)
Rho2=np.array([-1,-1],float)*math.floor(mSize/2)
# xCor,yCor=cv.polarToCart(Rho,Theta)
x=Rho2*np.cos(Theta)+math.ceil(mSize/2)
y=Rho2*np.sin(Theta)+math.ceil(mSize/2)
Mask2=np.zeros((mSize,mSize),np.float)
polyVerticesTemp=np.array(np.round([[x[0],y[0]],[x[1],y[1]],[mSize/2,mSize/2]]),np.int32)
# polyVertices=polyVerticesTemp.reshape(2,3)
# polyVerticesNew=polyVertices.transpose()
Mask2=cv.fillConvexPoly(Mask2, polyVerticesTemp, 1)
Mask=sc.logical_or(Mask1, Mask2)
Mask=Mask.astype(float)
plt.imshow(Mask)
N=np.float(cv.countNonZero(Mask))
plt.title(N)
plt.show()
else:
#rectangle generation
rho = np.array([1,1,-1,-1,1],float)*(mSize/2)
X,Y=cv.polarToCart(rho,theta)
#X=np.cos(theta)*rho
#Y=np.sin(theta)*rho
# X=X+math.ceil(mSize/2)
# Y=Y+math.ceil(mSize/2)
X=np.round(X+mSize/2)
Y=np.round(Y+mSize/2)
Mask=np.zeros((mSize,mSize),np.float)
polyVerticesTemp=np.array([X,Y],np.int32)
polyVertices=polyVerticesTemp.reshape(2,5)
polyVerticesNew=polyVertices.transpose()
Mask=cv.fillConvexPoly(Mask, polyVerticesNew, 1)
plt.imshow(Mask)
N=np.float(cv.countNonZero(Mask))
plt.title(N)
plt.show()
filts.append(Mask/N)
filts.append(Mask)
return filts, dirs
def get_limb_pts(eye_img, phi=20, angle_step=1, debug_index=False):
polar_img_w = 360 / angle_step # Polar image has one column per angle of interest
phi_range_1 = ((90 - phi) / angle_step, (90 + phi) / angle_step) # Ranges of angles to be ignored (too close to lids)
phi_range_2 = ((270 - phi) / angle_step, (270 + phi) / angle_step)
eye_img_grey = cv2.cvtColor(eye_img, cv2.COLOR_BGR2GRAY) # Do BGR-grey
eye_img_grey = cv2.medianBlur(eye_img_grey, 5)
# Scale to fixed size image for re-using transform matrix
scale = eye_img.shape[0] / float(__fixed_width)
img_fixed_size = cv2.resize(eye_img_grey, (__fixed_width, __fixed_width))
# Transform image into polar coords and blur
img_polar = linpolar(img_fixed_size, trans_w=polar_img_w, trans_h=__fixed_width / 2)
img_polar = cv2.GaussianBlur(img_polar, (5, 5), 0)
# Take the segment between min & max radii and filter with Gabor kernel
img_polar_seg = img_polar[__min_limb_r:__max_limb_r, :]
filter_img = cv2.filter2D(img_polar_seg, -1, __gabor_kern)
# Black out ignored angles
filter_img.T[ phi_range_1[0] : phi_range_1[1] ] = 0
filter_img.T[ phi_range_2[0] : phi_range_2[1] ] = 0
# In polar image, x <-> theta, y <-> magnitude
pol_ys = np.argmax(filter_img, axis=0) # Take highest filter response as limbus points
pol_xs = np.arange(filter_img.shape[1])[pol_ys > 0]
mags = (pol_ys + __min_limb_r)[pol_ys > 0]
thts = np.radians(pol_xs * angle_step)
# Translate each point back into fixed img coords
xs, ys = cv2.polarToCart(mags.astype(float), thts)
xs = (xs + __fixed_width / 2) * scale # Shift and scale cart. coords back to original eye-ROI coords
ys = (ys + __fixed_width / 2) * scale
# Points returned in form
# [[ x1 y1]
# [ x2 y2]
# ...
# [ xn yn]]
pts_cart = np.concatenate([xs, ys], axis=1)
# --------------------- Debug Drawing ---------------------
if debug_index != False:
debug_img = eye_img.copy()
debug_polar = cv2.cvtColor(img_polar, cv2.COLOR_GRAY2BGR)
cv2.imwrite("polar.jpg",debug_polar)
cv2.line(debug_polar, (0, __min_limb_r), (img_polar.shape[1], __min_limb_r), (255, 255, 0))
cv2.line(debug_polar, (0, __max_limb_r), (img_polar.shape[1], __max_limb_r), (255, 255, 0))
cv2.circle(debug_img, (debug_img.shape[1] / 2, debug_img.shape[0] / 2), int(debug_img.shape[0] * __limb_r_ratios[0]), (255, 255, 0))
cv2.circle(debug_img, (debug_img.shape[1] / 2, debug_img.shape[0] / 2), int(debug_img.shape[0] * __limb_r_ratios[1]), (255, 255, 0))
pts_polar = np.squeeze(np.dstack([pol_xs, mags]))
draw_points(debug_polar, pts_polar, (0, 0, 255), width=1)
draw_points(debug_img, pts_cart, (0, 0, 255), width=1)
stacked_imgs_polar = stack_imgs_vertical([debug_polar, filter_img])
stacked_imgs = stack_imgs_horizontal([debug_img, eye_img_grey, stacked_imgs_polar])
__debug_imgs[debug_index] = stacked_imgs
if debug_index == 2:
full_debug_img = stack_imgs_vertical([__debug_imgs[1], __debug_imgs[2]]);
cv2.imshow(__winname, full_debug_img)
elif debug_index > 2:
cv2.imshow(__winname, stacked_imgs);
# --------------------- Debug Drawing ---------------------
return pts_cart
def normalize_rect(rect):
pol_sq = np.array([[d[0][0] for d in cv2.cartToPolar(float(x), float(y))] for x, y in rect])
sorted_indices = np.lexsort((pol_sq[:, 1], pol_sq[:, 0]))
pol_sq = pol_sq[sorted_indices]
ordered_square = np.array([[d[0][0] for d in cv2.polarToCart(m, a)] for m, a in pol_sq], dtype=np.float32)
return ordered_square
img1 = cv2.imread("../hand.JPG", 0)
img2 = cv2.imread("../calender.JPG", 0)
imgf1 = np.float32(img1)
imgf2 = np.float32(img2)
dft1 = cv2.dft(imgf1, flags=cv2.DFT_COMPLEX_OUTPUT)
dft2 = cv2.dft(imgf2, flags=cv2.DFT_COMPLEX_OUTPUT)
# dftc1 = dft1
# dftc1[:,:,0] = dftc.real
# dftc1[:,:,1] = dftc.imag
dftmag1, dftphase1 = cv2.cartToPolar(dft1[:, :, 0], dft1[:, :, 1])
dftmag2, dftphase2 = cv2.cartToPolar(dft2[:, :, 0], dft2[:, :, 1])
dft1[:, :, 0], dft1[:, :, 1] = cv2.polarToCart(dftmag1, dftphase2)
dft2[:, :, 0], dft2[:, :, 1] = cv2.polarToCart(dftmag2, dftphase1)
# print dft1.shape
# dft_shift1 = np.fft.fftshift(dft1)
# dft_shift2 = np.fft.fftshift(dft2)
# magnitude_spectrum = 20*np.log(cv2.magnitude(dft_shift[:,:,0],dft_shift[:,:,1]))
# phase_spectrum = (cv2.phase(dft_shift[:,:,0],dft_shift[:,:,1]))
# dft_shift = np.fft.fftshift(dft1)
# magnitude_spectrum1 = 20*np.log(cv2.magnitude(dft_shift[:,:,0],dft_shift[:,:,1]))
# phase_spectrum1 = (cv2.phase(dft_shift[:,:,0],dft_shift[:,:,1]))
