I've read the documentation and searched Stack Overflow for the answer to this question, but can't find it. Sorry if it has already been answered.
I'm working with the results of an np.fft.fft2(Z) where Z is some 2d NumPy array. I would expect positive frequencies to be stored in values less than the Nyquist wavenumber in both x and y directions. From my tests, it seems this is the approach Matlab takes. In NumPy documentation they write positive frequencies are stored below the Nyquist number and negative frequencies above; this does not seem to be the case for fft2.
Some positive frequencies terms are stored at locations greater than the Nyquist wavenumber. For example, a mode at location (127,1) with associated amplitude stored at (1,127), will produce a 2D sinusoid with 4 peaks indicating that the wavenumber should be around 4, not 127.
I can't tell which is the positive and negative frequency in my example above because they are not following standard ordering.
So the main question I have is what kind of order does the fft2 follow for storing positive and negative frequencies?
I didn't post any examples because my question is a universal one and shouldn't be problem specific.
import numpy as np
from heapq import nlargest
## Setting up a simple example
lx = 4.0
ly = 4.0;
lz = 1.5;
nx = 128;
ny = 128;
L = 1.0
H = .4
x = np.linspace(0, lx, nx);
y = np.linspace(0, ly, ny);
x0 = 2.0;
y0 = 2.0;
z1 = np.zeros([ny,nx])
zm= np.zeros([ny,nx])
for j in range(1,ny):
for i in range(1,nx):
if np.sqrt(abs(x[i] - x0)** 2 + abs(y[j] - y0) ** 2) < L:
if np.abs(x[i] - x0) < L:
z1[j, i] = H * np.cos(np.pi * abs(x[i] - x0) / (2 *L))**2;
z1 = z1+np.transpose(z1)/2.0
## Here I take the fft
nf = np.shape(z1)[0]/2
fz1 = np.fft.fft2(z1)
spec_fz1 = np.abs(fz1)**2
valmax = nlargest(1000, spec_fz1.flatten())
## Here I search for amplitude pairs above nyquist number
for i in range(1,len(valmax),2):
xy = return_xy(valmax[i], spec_fz1)
if len(xy) >2:
if ((xy[0] > nf or xy[1]> nf) and (xy[2] > nf or xy[3]> nf) ):
print('both index locations above nyquist frequency')
else:
xy2 = return_xy(valmax[i+1], spec_fz1)
if ((xy[0] > nf or xy[1]> nf) and (xy2[0] > nf or xy2[1]> nf) ):
print('both index locations above nyquist frequency')
def return_xy(mode,spec_topo):
kxky = np.array([])
for i in range(np.shape(spec_topo)[0]):
for j in range(np.shape(spec_topo)[1]):
if spec_topo[i,j] == mode:
kxky= np.append(kxky,[i,j])
if len(kxky)> 1:
return kxky
else:
return kxky[0]
After sorting by the largest amplitude at the 21st index two amplitude pairs are stored at (127,1) and (1,127) which is above the Nyquist number. How should I interpret this wavenumber? note return_xy does same thing as np.where
I think this bit of code demonstrates how the 2D DFT output of np.fft.fft2 is organized:
import numpy as np
import matplotlib.pyplot as plt
n = 16
x = np.arange(n) / n * 2 * np.pi
y = np.arange(n) / n * 2 * np.pi
for kx in range(4):
for ky in range(4):
f = np.cos(kx * x[None,:] + ky * y[:,None])
F = np.fft.fft2(f)
plt.subplot(4, 4, 1 + ky * 4 + kx)
plt.imshow(np.abs(F))
plt.axis('off')
plt.title(f'kx = {kx}, ky = {ky}', fontsize=10)
plt.tight_layout()
plt.show()
We can see that the origin, kx=0 and ky=0 is at the top-left of the array. For a horizontal wave with exactly one period in the input, we see we have a pair of peaks at kx=1 and kx=N-1 (which is equivalent to kx=-1). With two periods in the input, kx=2 and kx=-2, etc. Vertical waves produce the same result but along the vertical axis, and diagonal waves at 45 degrees have the peaks at 45 degrees.
This is the exact same ordering as the 1D DFT (np.fft.fft) produces. The 2D DFT is simply the 1D DFT applied along the columns, and then along the rows of the result (or the other way around, it doesn't matter).
As for the test shown in the question, it is the superposition of two sine waves (one horizontal and one vertical) multiplied by a round window (a "pillbox" function). In the Fourier domain (continuous world), this corresponds to four impulse functions (two along the horizontal axis for the one sine wave, two along the vertical axis for the other sine wave), convolved with the Bessel function of the first kind of order 1 (J1). Because the sine waves have a low frequency, the four impulse functions are close together, and after the convolution appear as a somewhat wider Bessel function, centered around the origin:
plt.imshow(np.log(np.abs(fz1) + 1e-6))
plt.show()
What we see is the peak centered on the origin (at the top-left corner), with things to the left of the origin wrapped around to the right edge, and things to the top of the origin wrapped around to the bottom edge. Applying np.fft.fftshift moves the origin to the middle of the array, yielding a more recognizable shape.
Note: This question is building on another question of mine:
Two dimensional FFT using python results in slightly shifted frequency
I have some data, basically a function E(x,y) with (x,y) being a (discrete) subset of R^2, mapping to real numbers. For the (x,y) plane i have a fixed distance between data points in x- as well as in y direction (0,2). I want to analyze the frequency spectrum of my E(x,y) signal using a two dimensional fast fourier transform (FFT) using python.
As far as i know, no matter which frequencies are actually contained in my signal, using FFT, i will only be able to see signals below the Nyquisit limit Ny, which is Ny = sampling frequency / 2. In my case i have a real spacing of 0,2, leading to a sampling frequency of 1 / 0,2 = 5 and therefore my Nyquisit limit is Ny = 5 / 2 = 2,5.
If my signal does have frequencies above the Nyquisit limit, they will be "folded" back into the Nyquisit domain, leading to false results (aliasing). But even though i might sample with a too low frequency, it should in theory not be possible to see any frequencies above the Niquisit limit, correct?
So here is my issue: Analyzing my signal should only lead to frequencies of 2,5 max., but i cleary get frequencies higher than that. Given that i am pretty sure about the theory here, there has to be some mistake in my code. I will provide a shortened code version, only providing necessary information for this issue:
simulationArea =... # length of simulation area in x and y direction
x = np.linspace(0, simulationArea, numberOfGridPointsInX, endpoint=False)
y = x
xx, yy = np.meshgrid(x, y)
Ex = np.genfromtxt('E_field_x100.txt') # this is the actual signal to be analyzed, which may have arbitrary frequencies
FTEx = np.fft.fft2(Ex) # calculating fft coefficients of signal
dx = x[1] - x[0] # calculating spacing of signals in real space. 'print(dx)' results in '0.2'
sampleFrequency = 1.0 / dx
nyquisitFrequency = sampleFrequency / 2.0
half = len(FTEx) / 2
fig, axarr = plt.subplots(2, 1)
im1 = axarr[0, 0].imshow(Ex,
origin='lower',
cmap='jet',
extent=(0, simulationArea, 0, simulationArea))
axarr[0, 0].set_xlabel('X', fontsize=14)
axarr[0, 0].set_ylabel('Y', fontsize=14)
axarr[0, 0].set_title('$E_x$', fontsize=14)
fig.colorbar(im1, ax=axarr[0, 0])
im2 = axarr[1, 0].matshow(2 * abs(FTEx[:half, :half]) / half,
aspect='equal',
origin='lower',
interpolation='nearest')
axarr[1, 0].set_xlabel('Frequency wx')
axarr[1, 0].set_ylabel('Frequency wy')
axarr[1, 0].xaxis.set_ticks_position('bottom')
axarr[1, 0].set_title('$FFT(E_x)$', fontsize=14)
fig.colorbar(im2, ax=axarr[1, 0])
The result of this is:
How is that possible? When i am using the same code for very simple signals, it works just fine (e.g. a sine wave in x or y direction with a specific frequency).
Ok here we go! Here’s a couple of simple functions and a complete example that you can use: it’s got a little bit of extra cruft related to plotting and for data generation but the first function, makeSpectrum shows you how to use fftfreq and fftshift plus fft2 to achieve what you want. Let me know if you have questions.
import numpy as np
import numpy.fft as fft
import matplotlib.pylab as plt
def makeSpectrum(E, dx, dy, upsample=10):
"""
Convert a time-domain array `E` to the frequency domain via 2D FFT. `dx` and
`dy` are sample spacing in x (left-right, 1st axis) and y (up-down, 0th
axis) directions. An optional `upsample > 1` will zero-pad `E` to obtain an
upsampled spectrum.
Returns `(spectrum, xf, yf)` where `spectrum` contains the 2D FFT of `E`. If
`Ny, Nx = spectrum.shape`, `xf` and `yf` will be vectors of length `Nx` and
`Ny` respectively, containing the frequencies corresponding to each pixel of
`spectrum`.
The returned spectrum is zero-centered (via `fftshift`). The 2D FFT, and
this function, assume your input `E` has its origin at the top-left of the
array. If this is not the case, i.e., your input `E`'s origin is translated
away from the first pixel, the returned `spectrum`'s phase will *not* match
what you expect, since a translation in the time domain is a modulation of
the frequency domain. (If you don't care about the spectrum's phase, i.e.,
only magnitude, then you can ignore all these origin issues.)
"""
zeropadded = np.array(E.shape) * upsample
F = fft.fftshift(fft.fft2(E, zeropadded)) / E.size
xf = fft.fftshift(fft.fftfreq(zeropadded[1], d=dx))
yf = fft.fftshift(fft.fftfreq(zeropadded[0], d=dy))
return (F, xf, yf)
def extents(f):
"Convert a vector into the 2-element extents vector imshow needs"
delta = f[1] - f[0]
return [f[0] - delta / 2, f[-1] + delta / 2]
def plotSpectrum(F, xf, yf):
"Plot a spectrum array and vectors of x and y frequency spacings"
plt.figure()
plt.imshow(abs(F),
aspect="equal",
interpolation="none",
origin="lower",
extent=extents(xf) + extents(yf))
plt.colorbar()
plt.xlabel('f_x (Hz)')
plt.ylabel('f_y (Hz)')
plt.title('|Spectrum|')
plt.show()
if __name__ == '__main__':
# In seconds
x = np.linspace(0, 4, 20)
y = np.linspace(0, 4, 30)
# Uncomment the next two lines and notice that the spectral peak is no
# longer equal to 1.0! That's because `makeSpectrum` expects its input's
# origin to be at the top-left pixel, which isn't the case for the following
# two lines.
# x = np.linspace(.123 + 0, .123 + 4, 20)
# y = np.linspace(.123 + 0, .123 + 4, 30)
# Sinusoid frequency, in Hz
x0 = 1.9
y0 = -2.9
# Generate data
im = np.exp(2j * np.pi * (y[:, np.newaxis] * y0 + x[np.newaxis, :] * x0))
# Generate spectrum and plot
spectrum, xf, yf = makeSpectrum(im, x[1] - x[0], y[1] - y[0])
plotSpectrum(spectrum, xf, yf)
# Report peak
peak = spectrum[:, np.isclose(xf, x0)][np.isclose(yf, y0)]
peak = peak[0, 0]
print('spectral peak={}'.format(peak))
Results in the following image, and prints out, spectral peak=(1+7.660797103157986e-16j), which is exactly the correct value for the spectrum at the frequency of a pure complex exponential.
I'm interested in solving,
\frac{\delta \phi}{\delta t} - D \nabla^2 \phi - \alpha \phi - \gamma \phi = 0
The following is working, but I have a few questions:
Is it possible to increase performance with FiPy? I feel like the nx, ny, nz bins are very small here, despite a long computation time. I don't understand why the arrays X, Y, and Z are so large.
Notice in the first frame, we are zoomed in. How can I force the extents to automatically be [0..nx, 0..ny, 0..nz] in all plots?
Data for the first frame is a sphere of points with values 1.0 surrounded by 0.0. Why does there appear to be a gradient? Is Mayavi interpolating? If so, how can I disable this?
Code:
from fipy import *
import mayavi.mlab as mlab
import numpy as np
import time
# Spatial parameters
nx = ny = nz = 30 # bins
dx = dy = dz = 1 # Must this be an integer?
L = nx * dx
# Diffusion and time step
D = 1.
dt = 10.0 * dx**2 / (2. * D)
steps = 4
# Initial value and radius of concentration
phi0 = 1.0
r = 3.0
# Rates
alpha = 1.0 # Source coeficcient
gamma = .01 # Sink coeficcient
mesh = Grid3D(nx=nx, ny=ny, nz=nz, dx=dx, dy=dy, dz=dz)
X, Y, Z = mesh.cellCenters # These are large arrays
phi = CellVariable(mesh=mesh, name=r"$\phi$", value=0.)
src = phi * alpha # Source term (zeroth order reaction)
degr = -gamma * phi # Sink term (degredation)
eq = TransientTerm() == DiffusionTerm(D) + src + degr
# Initial concentration is a sphere located in the center of a bounded cube
phi.setValue(1.0, where=( ((X-nx/2))**2 + (Y-ny/2)**2 + (Z-nz/2)**2 < r**2) )
# Solve
start_time = time.time()
results = [phi.getNumericValue().copy()]
for step in range(steps):
eq.solve(var=phi, dt=dt)
results.append(phi.getNumericValue().copy())
print 'Time elapsed:', time.time() - start_time
# Plot
for i, res in enumerate(results):
fig = mlab.figure()
res = res.reshape(nx, ny, nz)
mlab.contour3d(res, opacity=.3, vmin=0, vmax=1, contours=100, transparent=True, extent=[0, 10, 0, 10, 0, 10])
mlab.colorbar()
mlab.savefig('diffusion3d_%i.png'%(i+1))
mlab.close()
Time elapsed: 68.2 seconds
It's hard to tell from your question, but in the course of diagnosing things, I discovered that the LinearLUSolver scales very poorly as the dimension of the problem increases (see https://github.com/usnistgov/fipy/issues/474).
For this symmetric problem, PySparse should use the PCG solver and Trilinos should use GMRES. If you didn't install either of these, then you'll get the SciPy sparse solvers, which defaults to LU (I don't know why; something for us to look into), and things will be really slow in 3D. Try adding solver=LinearGMRESSolver() to your eq.solve(...) statement.
As far as the size of X, Y, and Z, you've declared a 30*30*30 cube of cells, so each of the cell center coordinate vectors will be 27000 elements long. Did you have a different expectation for cellCenters?
I suggest you subclass our MayaviDaemon class, or at least look at how it sets up the display in Mayavi. In short, we set a data_set_clipper to the desired bounds.
I don't know.
I would like to use the Fourier transform to find the center of a simulated entity under periodic boundary condition; periodic boundary conditions means, that whenever something exits through one side of the box, it is warped around to appear on the opposite side just like in the classic game asteroids.
So what I have is for each time frame a matrix (Nx3) with N the number of points in xyz. what I want to do is determine the center of that cloud even if it all moved over the periodic boundary and is so to say stuck in between.
My idea for an solution would now be do a (mass weigted) histogram of these points and then perform an FFT on that and use the phase of the first Fourier coefficient to determine where in the box the maximum would be.
as a test case I have used
import numpy as np
Points_x = np.random.randn(10000)
Box_min = -10
Box_max = 10
X = np.linspace( Box_min, Box_max, 100 )
### make a Histogram of the points
Histogram_Points = np.bincount( np.digitize( Points_x, X ), minlength=100 )
### make an artifical shift over the periodic boundary
Histogram_Points = np.r_[ Histogram_Points[45:], Histogram_Points[:45] ]
So now I can use FFT since it expects a periodic function anyways.
## doing fft
F = np.fft.fft(Histogram_Points)
## getting rid of everything but first harmonic
F[2:] = 0.
## back transforming
Fist_harmonic = np.fft.ifft(F)
That way I get a sine wave with its maximum exactly where the maximum of the histogram is.
Now I'd like to extract the position of the maximum not by taking the max function on the sine vector, but somehow it should be retrievable from the first (not the 0th) Fourier coefficient, since that should somehow contain the phase shift of the sine to have its maximum exactly at the maximum of the histogram.
Indeed, plotting
Cos_approx = cos( linspace(0,2*pi,100) * angle(F[1]) )
will give
But I can't figure out how to get the position of the peak from this angle.
Using the FFT is overkill when all you need is one Fourier coefficent. Instead, you can simply compute the dot product of your data with
w = np.exp(-2j*np.pi*np.arange(N) / N)
where N is the number of points. (The time to compute all the Fourier coefficients with the FFT is O(N*log(N)). Computing just one coefficient is O(N).)
Here's a script similar to yours. The data is put in y; the coordinates of the data points are in x.
import numpy as np
N = 100
# x coordinates of the data
xmin = -10
xmax = 10
x = np.linspace(xmin, xmax, N, endpoint=False)
# Generate data in y.
n = 35
y = np.zeros(N)
y[:n] = 1 - np.cos(np.linspace(0, 2*np.pi, n))
y[:n] /= 0.7 + 0.3*np.random.rand(n)
m = 10
y = np.r_[y[m:], y[:m]]
# Compute coefficent 1 of the discrete Fourier transform.
w = np.exp(-2j*np.pi*np.arange(N) / N)
F1 = y.dot(w)
print "F1 =", F1
# Get the angle of F1 (in the interval [0,2*pi]).
angle = np.angle(F1.conj())
if angle < 0:
angle += 2*np.pi
center_x = xmin + (xmax - xmin) * angle / (2*np.pi)
print "center_x = ", center_x
# Create the first sinusoidal mode for the plot.
mode1 = (F1.real * np.cos(2*np.pi*np.arange(N)/N) -
F1.imag*np.sin(2*np.pi*np.arange(N)/N))/np.abs(F1)
import matplotlib.pyplot as plt
plt.clf()
plt.plot(x, y)
plt.plot(x, mode1)
plt.axvline(center_x, color='r', linewidth=1)
plt.show()
This generates the plot:
To answer the question "Why F1.conj()?":
The complex conjugate of F1 is used because of the minus sign in
w = np.exp(-2j*np.pi*np.arange(N) / N) (which I used because it
is a common convention).
Since w can be written
w = np.exp(-2j*np.pi*np.arange(N) / N)
= cos(-2*pi*arange(N)/N) + 1j*sin(-2*pi*arange(N)/N)
= cos(2*pi*arange(N)/N) - 1j*sin(2*pi*arange(N)/N)
the dot product y.dot(w) is basically a projection of y onto
cos(2*pi*arange(N)/N) (the real part of F1) and -sin(2*pi*arange(N)/N)
(the imaginary part of F1). But when we figure out the phase of
the maximum, it is based on the functions cos(...) and sin(...). Taking
the complex conjugate accounts for the opposite sign of the sin()
function. If w = np.exp(2j*np.pi*np.arange(N) / N) were used instead, the
complex conjugate of F1 would not be needed.
You could calculate the circular mean directly on your data.
When calculating the circular mean, your data is mapped to -pi..pi. This mapped data is interpreted as angle to a point on the unit circle. Then the mean value of x and y component is calculated. The next step is to calculate the resulting angle and map it back to the defined "box".
import numpy as np
import matplotlib.pyplot as plt
Points_x = np.random.randn(10000)+1
Box_min = -10
Box_max = 10
Box_width = Box_max - Box_min
#Maps Points to Box_min ... Box_max with periodic boundaries
Points_x = (Points_x%Box_width + Box_min)
#Map Points to -pi..pi
Points_map = (Points_x - Box_min)/Box_width*2*np.pi-np.pi
#Calc circular mean
Pmean_map = np.arctan2(np.sin(Points_map).mean() , np.cos(Points_map).mean())
#Map back
Pmean = (Pmean_map+np.pi)/(2*np.pi) * Box_width + Box_min
#Plotting the result
plt.figure(figsize=(10,3))
plt.subplot(121)
plt.hist(Points_x, 100);
plt.plot([Pmean, Pmean], [0, 1000], c='r', lw=3, alpha=0.5);
plt.subplot(122,aspect='equal')
plt.plot(np.cos(Points_map), np.sin(Points_map), '.');
plt.ylim([-1, 1])
plt.xlim([-1, 1])
plt.grid()
plt.plot([0, np.cos(Pmean_map)], [0, np.sin(Pmean_map)], c='r', lw=3, alpha=0.5);
I am trying to use circle fitting code for 3D data set. I have modified it for 3D points just adding z-coordinate where necessary. My modification works fine for one set of points and works bad for another. Please look at the code, if it has some errors.
import trig_items
import numpy as np
from trig_items import *
from numpy import *
from matplotlib import pyplot as p
from scipy import optimize
# Coordinates of the 3D points
##x = r_[36, 36, 19, 18, 33, 26]
##y = r_[14, 10, 28, 31, 18, 26]
##z = r_[0, 1, 2, 3, 4, 5]
x = r_[ 2144.18908574, 2144.26880854, 2144.05552972, 2143.90303742, 2143.62520676,
2143.43628579, 2143.14005775, 2142.79919654, 2142.51436023, 2142.11240866,
2141.68564346, 2141.29333828, 2140.92596405, 2140.3475612, 2139.90848046,
2139.24661021, 2138.67384709, 2138.03313547, 2137.40301734, 2137.40908256,
2137.06611224, 2136.50943781, 2136.0553113, 2135.50313189, 2135.07049922,
2134.62098139, 2134.10459535, 2133.50838433, 2130.6600465, 2130.03537342,
2130.04047644, 2128.83522468, 2127.79827542, 2126.43513385, 2125.36700593,
2124.00350543, 2122.68564431, 2121.20709478, 2119.79047011, 2118.38417647,
2116.90063343, 2115.52685778, 2113.82246629, 2112.21159431, 2110.63180117,
2109.00713198, 2108.94434529, 2106.82777156, 2100.62343757, 2098.5090226,
2096.28787738, 2093.91550703, 2091.66075061, 2089.15316429, 2086.69753869,
2084.3002414, 2081.87590579, 2079.19141866, 2076.5394574, 2073.89128676,
2071.18786213]
y = r_[ 725.74913818, 724.43874065, 723.15226506, 720.45950581, 717.77827954,
715.07048092, 712.39633862, 709.73267688, 707.06039438, 704.43405908,
701.80074596, 699.15371526, 696.5309022, 693.96109921, 691.35585501,
688.83496327, 686.32148661, 683.80286662, 681.30705568, 681.30530975,
679.66483676, 678.01922321, 676.32721779, 674.6667554, 672.9658024,
671.23686095, 669.52021535, 667.84999077, 659.19757984, 657.46179949,
657.45700508, 654.46901086, 651.38177517, 648.41739432, 645.32356976,
642.39034578, 639.42628453, 636.51107198, 633.57732055, 630.63825133,
627.75308356, 624.80162215, 622.01980232, 619.18814892, 616.37688894,
613.57400131, 613.61535723, 610.4724493, 600.98277781, 597.84782844,
594.75983001, 591.77946964, 588.74874068, 585.84525834, 582.92311166,
579.99564481, 577.06666417, 574.30782762, 571.54115037, 568.79760614,
566.08551098]
z = r_[ 339.77146775, 339.60021095, 339.47645894, 339.47130963, 339.37216218,
339.4126132, 339.67942046, 339.40917728, 339.39500353, 339.15041461,
339.38959195, 339.3358209, 339.47764895, 339.17854867, 339.14624071,
339.16403926, 339.02308811, 339.27011082, 338.97684183, 338.95087698,
338.97321177, 339.02175448, 339.02543922, 338.88725411, 339.06942374,
339.0557553, 339.04414618, 338.89234303, 338.95572249, 339.00880416,
339.00413073, 338.91080374, 338.98214758, 339.01135789, 338.96393537,
338.73446188, 338.62784913, 338.72443217, 338.74880562, 338.69090173,
338.50765186, 338.49056867, 338.57353355, 338.6196255, 338.43754399,
338.27218569, 338.10587265, 338.43880881, 338.28962141, 338.14338705,
338.25784154, 338.49792568, 338.15572139, 338.52967693, 338.4594245,
338.1511823, 338.03711207, 338.19144663, 338.22022045, 338.29032321,
337.8623197 ]
# coordinates of the barycenter
xm = mean(x)
ym = mean(y)
zm = mean(z)
### Basic usage of optimize.leastsq
def calc_R(xc, yc, zc):
""" calculate the distance of each 3D points from the center (xc, yc, zc) """
return sqrt((x - xc) ** 2 + (y - yc) ** 2 + (z - zc) ** 2)
def func(c):
""" calculate the algebraic distance between the 3D points and the mean circle centered at c=(xc, yc, zc) """
Ri = calc_R(*c)
return Ri - Ri.mean()
center_estimate = xm, ym, zm
center, ier = optimize.leastsq(func, center_estimate)
##print center
xc, yc, zc = center
Ri = calc_R(xc, yc, zc)
R = Ri.mean()
residu = sum((Ri - R)**2)
print 'R =', R
So, for the first set of x, y, z (commented in the code) it works well: the output is R = 39.0097846735. If I run the code with the second set of points (uncommented) the resulting radius is R = 108576.859834, which is almost straight line. I plotted the last one.
The blue points is a given data set, the red ones is the arc of the resulting radius R = 108576.859834. It is obvious that the given data set has much smaller radius than the result.
Here is another set of points.
It is clear that the least squares does not work correctly.
Please help me solving this issue.
UPDATE
Here is my solution:
### fit 3D arc into a set of 3D points ###
### output is the centre and the radius of the arc ###
def fitArc3d(arr, eps = 0.0001):
# Coordinates of the 3D points
x = numpy.array([arr[k][0] for k in range(len(arr))])
y = numpy.array([arr[k][4] for k in range(len(arr))])
z = numpy.array([arr[k][5] for k in range(len(arr))])
# coordinates of the barycenter
xm = mean(x)
ym = mean(y)
zm = mean(z)
### gradient descent minimisation method ###
pnts = [[x[k], y[k], z[k]] for k in range(len(x))]
meanP = Point(xm, ym, zm) # mean point
Ri = [Point(*meanP).distance(Point(*pnts[k])) for k in range(len(pnts))] # radii to the points
Rm = math.fsum(Ri) / len(Ri) # mean radius
dR = Rm + 10 # difference between mean radii
alpha = 0.1
c = meanP
cArr = []
while dR > eps:
cArr.append(c)
Jx = math.fsum([2 * (x[k] - c[0]) * (Ri[k] - Rm) / Ri[k] for k in range(len(Ri))])
Jy = math.fsum([2 * (y[k] - c[1]) * (Ri[k] - Rm) / Ri[k] for k in range(len(Ri))])
Jz = math.fsum([2 * (z[k] - c[2]) * (Ri[k] - Rm) / Ri[k] for k in range(len(Ri))])
gradJ = [Jx, Jy, Jz] # find gradient
c = [c[k] + alpha * gradJ[k] for k in range(len(c)) if len(c) == len(gradJ)] # find new centre point
Ri = [Point(*c).distance(Point(*pnts[k])) for k in range(len(pnts))] # calculate new radii
RmOld = Rm
Rm = math.fsum(Ri) / len(Ri) # calculate new mean radius
dR = abs(Rm - RmOld) # new difference between mean radii
return Point(*c), Rm
It is not very optimal code (I do not have time to fine tune it) but it works.
I guess the problem is the data and the corresponding algorithm. The least square method works fine if it produces a local parabolic minimum, such that a simple gradient method goes approximately direction minimum. Unfortunately, this is not necessarily the case for your data. You can check this by keeping some rough estimates for xc and yc fixed and plotting the sum of the squared residuals as a function of zc and R. I get a boomerang shaped minimum. Depending on your starting parameters you might end in one of the branches going away from the real minimum. Once in the valley this can be very flat such that you exceed the number of max iterations or get something that is accepted within the tolerance of the algorithm. As always, thinks are better the better your starting parameters. Unfortunately you have only a small arc of the circle, so that it is difficult to get better. I am not a specialist in Python, but I think that leastsq allows you to play with the Jacobian and Gradient Methods. Try to play with the tolerance as well.
In short: the code looks basically fine to me, but your data is pathological and you have to adapt the code to that kind of data.
There is a non-iterative solution in 2D from Karimäki, maybe you can adapt
this method to 3D. You can also look at this. Sure you will find more literature.
I just checked the data using a Simplex-Algorithm. The minimum is, as I said, not well behaved. See here some cuts of the residual function. Only in the xy-plane you get some reasonable behavior. The properties of the zr- and xr- plane make the finding process very difficult.
So in the beginning the simplex algorithm finds several almost stable solutions. You can see them as flat steps in the graph below (blue x, purple y, yellow z, green R). At the end the algorithm has to walk down the almost flat but very stretched out valley, resulting in the final conversion of z and R. Nevertheless, I expect many regions that look like a solution if the tolerance is insufficient. With the standard tolerance of 10^-5 the algoritm stopped after approx 350 iterations. I had to set it to 10^-10 to get this solution, i.e. [1899.32, 741.874, 298.696, 248.956], which seems quite ok.
Update
As mentioned earlier, the solution depends on the working precision and requested accuracy. So your hand made gradient method works probably better as these values are different compared to the build-in least square fit. Nevertheless, this is my version making a two step fit. First I fit a plane to the data. In a next step I fit a circle within this plane. Both steps use the least square method. This time it works, as each step avoids critically shaped minima. (Naturally, the plane fit runs into problems if the arc segment becomes small and the data lies virtually on a straight line. But this will happen for all algorithms)
from math import *
from matplotlib import pyplot as plt
from scipy import optimize
import numpy as np
from mpl_toolkits.mplot3d import Axes3D
import pprint as pp
dataTupel=zip(xs,ys,zs) #your data from above
# Fitting a plane first
# let the affine plane be defined by two vectors,
# the zero point P0 and the plane normal n0
# a point p is member of the plane if (p-p0).n0 = 0
def distanceToPlane(p0,n0,p):
return np.dot(np.array(n0),np.array(p)-np.array(p0))
def residualsPlane(parameters,dataPoint):
px,py,pz,theta,phi = parameters
nx,ny,nz =sin(theta)*cos(phi),sin(theta)*sin(phi),cos(theta)
distances = [distanceToPlane([px,py,pz],[nx,ny,nz],[x,y,z]) for x,y,z in dataPoint]
return distances
estimate = [1900, 700, 335,0,0] # px,py,pz and zeta, phi
#you may automize this by using the center of mass data
# note that the normal vector is given in polar coordinates
bestFitValues, ier = optimize.leastsq(residualsPlane, estimate, args=(dataTupel))
xF,yF,zF,tF,pF = bestFitValues
point = [xF,yF,zF]
normal = [sin(tF)*cos(pF),sin(tF)*sin(pF),cos(tF)]
# Fitting a circle inside the plane
#creating two inplane vectors
sArr=np.cross(np.array([1,0,0]),np.array(normal))#assuming that normal not parallel x!
sArr=sArr/np.linalg.norm(sArr)
rArr=np.cross(sArr,np.array(normal))
rArr=rArr/np.linalg.norm(rArr)#should be normalized already, but anyhow
def residualsCircle(parameters,dataPoint):
r,s,Ri = parameters
planePointArr = s*sArr + r*rArr + np.array(point)
distance = [ np.linalg.norm( planePointArr-np.array([x,y,z])) for x,y,z in dataPoint]
res = [(Ri-dist) for dist in distance]
return res
estimateCircle = [0, 0, 335] # px,py,pz and zeta, phi
bestCircleFitValues, ier = optimize.leastsq(residualsCircle, estimateCircle,args=(dataTupel))
rF,sF,RiF = bestCircleFitValues
print bestCircleFitValues
# Synthetic Data
centerPointArr=sF*sArr + rF*rArr + np.array(point)
synthetic=[list(centerPointArr+ RiF*cos(phi)*rArr+RiF*sin(phi)*sArr) for phi in np.linspace(0, 2*pi,50)]
[cxTupel,cyTupel,czTupel]=[ x for x in zip(*synthetic)]
### Plotting
d = -np.dot(np.array(point),np.array(normal))# dot product
# create x,y mesh
xx, yy = np.meshgrid(np.linspace(2000,2200,10), np.linspace(540,740,10))
# calculate corresponding z
# Note: does not work if normal vector is without z-component
z = (-normal[0]*xx - normal[1]*yy - d)/normal[2]
# plot the surface, data, and synthetic circle
fig = plt.figure()
ax = fig.add_subplot(211, projection='3d')
ax.scatter(xs, ys, zs, c='b', marker='o')
ax.plot_wireframe(xx,yy,z)
ax.set_xlabel('X Label')
ax.set_ylabel('Y Label')
ax.set_zlabel('Z Label')
bx = fig.add_subplot(212, projection='3d')
bx.scatter(xs, ys, zs, c='b', marker='o')
bx.scatter(cxTupel,cyTupel,czTupel, c='r', marker='o')
bx.set_xlabel('X Label')
bx.set_ylabel('Y Label')
bx.set_zlabel('Z Label')
plt.show()
which give a radius of 245. This is close to what the other approach gave (249). So within error margins I get the same.
The plotted result looks reasonable.
Hope this helps.
Feel like you missed some constraints in your 1st version code. The implementation could be explained as fitting a sphere to 3d points. So that's why the 2nd radius for 2nd data list is almost straight line. It's thinking like you are giving it a small circle on a large sphere.