We Keep Coding

Way to contour outer edge of selected grid region in Python

I have the following code:
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(-np.pi/2, np.pi/2, 30)
y = np.linspace(-np.pi/2, np.pi/2, 30)
x,y = np.meshgrid(x,y)
z = np.sin(x**2+y**2)[:-1,:-1]
fig,ax = plt.subplots()
ax.pcolormesh(x,y,z)
Which gives this image:
Now lets say I want to highlight the edge certain grid boxes:
highlight = (z > 0.9)
I could use the contour function, but this would result in a "smoothed" contour. I just want to highlight the edge of a region, following the edge of the grid boxes.
The closest I've come is adding something like this:
highlight = np.ma.masked_less(highlight, 1)
ax.pcolormesh(x, y, highlight, facecolor = 'None', edgecolors = 'w')
Which gives this plot:
Which is close, but what I really want is for only the outer and inner edges of that "donut" to be highlighted.
So essentially I am looking for some hybrid of the contour and pcolormesh functions - something that follows the contour of some value, but follows grid bins in "steps" rather than connecting point-to-point. Does that make sense?
Side note: In the pcolormesh arguments, I have edgecolors = 'w', but the edges still come out to be blue. Whats going on there?
EDIT:
JohanC's initial answer using add_iso_line() works for the question as posed. However, the actual data I'm using is a very irregular x,y grid, which cannot be converted to 1D (as is required for add_iso_line().
I am using data which has been converted from polar coordinates (rho, phi) to cartesian (x,y). The 2D solution posed by JohanC does not appear to work for the following case:
import numpy as np
import matplotlib.pyplot as plt
from scipy import ndimage
def pol2cart(rho, phi):
x = rho * np.cos(phi)
y = rho * np.sin(phi)
return(x, y)
phi = np.linspace(0,2*np.pi,30)
rho = np.linspace(0,2,30)
pp, rr = np.meshgrid(phi,rho)
xx,yy = pol2cart(rr, pp)
z = np.sin(xx**2 + yy**2)
scale = 5
zz = ndimage.zoom(z, scale, order=0)
fig,ax = plt.subplots()
ax.pcolormesh(xx,yy,z[:-1, :-1])
xlim = ax.get_xlim()
ylim = ax.get_ylim()
xmin, xmax = xx.min(), xx.max()
ymin, ymax = yy.min(), yy.max()
ax.contour(np.linspace(xmin,xmax, zz.shape[1]) + (xmax-xmin)/z.shape[1]/2,
np.linspace(ymin,ymax, zz.shape[0]) + (ymax-ymin)/z.shape[0]/2,
np.where(zz < 0.9, 0, 1), levels=[0.5], colors='red')
ax.set_xlim(*xlim)
ax.set_ylim(*ylim)

This post shows a way to draw such lines. As it is not straightforward to adapt to the current pcolormesh, the following code demonstrates a possible adaption.
Note that the 2d versions of x and y have been renamed, as the 1d versions are needed for the line segments.
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.collections import LineCollection
x = np.linspace(-np.pi / 2, np.pi / 2, 30)
y = np.linspace(-np.pi / 2, np.pi / 2, 30)
xx, yy = np.meshgrid(x, y)
z = np.sin(xx ** 2 + yy ** 2)[:-1, :-1]
fig, ax = plt.subplots()
ax.pcolormesh(x, y, z)
def add_iso_line(ax, value, color):
v = np.diff(z > value, axis=1)
h = np.diff(z > value, axis=0)
l = np.argwhere(v.T)
vlines = np.array(list(zip(np.stack((x[l[:, 0] + 1], y[l[:, 1]])).T,
np.stack((x[l[:, 0] + 1], y[l[:, 1] + 1])).T)))
l = np.argwhere(h.T)
hlines = np.array(list(zip(np.stack((x[l[:, 0]], y[l[:, 1] + 1])).T,
np.stack((x[l[:, 0] + 1], y[l[:, 1] + 1])).T)))
lines = np.vstack((vlines, hlines))
ax.add_collection(LineCollection(lines, lw=1, colors=color))
add_iso_line(ax, 0.9, 'r')
plt.show()
Here is an adaption of the second answer, which can work with only 2d arrays:
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.collections import LineCollection
from scipy import ndimage
x = np.linspace(-np.pi / 2, np.pi / 2, 30)
y = np.linspace(-np.pi / 2, np.pi / 2, 30)
x, y = np.meshgrid(x, y)
z = np.sin(x ** 2 + y ** 2)
scale = 5
zz = ndimage.zoom(z, scale, order=0)
fig, ax = plt.subplots()
ax.pcolormesh(x, y, z[:-1, :-1] )
xlim = ax.get_xlim()
ylim = ax.get_ylim()
xmin, xmax = x.min(), x.max()
ymin, ymax = y.min(), y.max()
ax.contour(np.linspace(xmin,xmax, zz.shape[1]) + (xmax-xmin)/z.shape[1]/2,
np.linspace(ymin,ymax, zz.shape[0]) + (ymax-ymin)/z.shape[0]/2,
np.where(zz < 0.9, 0, 1), levels=[0.5], colors='red')
ax.set_xlim(*xlim)
ax.set_ylim(*ylim)
plt.show()

I'll try to refactor add_iso_line method in order to make it more clear an open for optimisations. So, at first, there comes a must-do part:
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.collections import LineCollection
x = np.linspace(-np.pi/2, np.pi/2, 30)
y = np.linspace(-np.pi/2, np.pi/2, 30)
x, y = np.meshgrid(x,y)
z = np.sin(x**2+y**2)[:-1,:-1]
fig, ax = plt.subplots()
ax.pcolormesh(x,y,z)
xlim, ylim = ax.get_xlim(), ax.get_ylim()
highlight = (z > 0.9)
Now highlight is a binary array that looks like this:
After that we can extract indexes of True cells, look for False neighbourhoods and identify positions of 'red' lines. I'm not comfortable enough with doing it in a vectorised manner (like here in add_iso_line method) so just using simple loop:
lines = []
cells = zip(*np.where(highlight))
for x, y in cells:
if x == 0 or highlight[x - 1, y] == 0: lines.append(([x, y], [x, y + 1]))
if x == highlight.shape[0] or highlight[x + 1, y] == 0: lines.append(([x + 1, y], [x + 1, y + 1]))
if y == 0 or highlight[x, y - 1] == 0: lines.append(([x, y], [x + 1, y]))
if y == highlight.shape[1] or highlight[x, y + 1] == 0: lines.append(([x, y + 1], [x + 1, y + 1]))
And, finally, I resize and center coordinates of lines in order to fit with pcolormesh:
lines = (np.array(lines) / highlight.shape - [0.5, 0.5]) * [xlim[1] - xlim[0], ylim[1] - ylim[0]]
ax.add_collection(LineCollection(lines, colors='r'))
plt.show()
In conclusion, this is very similar to JohanC solution and, in general, slower. Fortunately, we can reduce amount of cells significantly, extracting contours only using python-opencv package:
import cv2
highlight = highlight.astype(np.uint8)
contours, hierarchy = cv2.findContours(highlight, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
cells = np.vstack(contours).squeeze()
This is an illustration of cells being checked:

Matplotlib contour hatching not working if only two levels was used

I am trying to plot hatches over contours lines that
statisfy certian criteria folliwng the example found here. Yet, I got regular contours (the yellow lines) instead of the hatches. Any ideas how to resolve that. Thanks
import matplotlib.pyplot as plt
import numpy as np
# invent some numbers, turning the x and y arrays into simple
# 2d arrays, which make combining them together easier.
x = np.linspace(-3, 5, 150).reshape(1, -1)
y = np.linspace(-3, 5, 120).reshape(-1, 1)
z = np.cos(x) + np.sin(y)
# we no longer need x and y to be 2 dimensional, so flatten them.
x, y = x.flatten(), y.flatten()
fig2, ax2 = plt.subplots()
n_levels = 6
a=ax2.contourf(x, y, z, n_levels)
fig2.colorbar(a)
[m,n]=np.where(z > 0.5)
z1=np.zeros(z.shape)
z1[m,n]=99
cs = ax2.contour(x, y, z1,2,hatches=['','.'])
plt.show()enter code here

Use contourf() with proper parameters to get useful plot with hatching. See important comment within the working code below:
import matplotlib.pyplot as plt
import numpy as np
x = np.linspace(-3, 5, 150).reshape(1, -1)
y = np.linspace(-3, 5, 120).reshape(-1, 1)
z = np.cos(x) + np.sin(y)
x, y = x.flatten(), y.flatten()
fig2, ax2 = plt.subplots()
n_levels = 6
a = ax2.contourf(x, y, z, n_levels)
fig2.colorbar(a)
[m,n] = np.where(z > 0.5)
z1=np.zeros(z.shape)
z1[m, n] = 99
# use contourf() with proper hatch pattern and alpha value
cs = ax2.contourf(x, y, z1 ,3 , hatches=['', '..'], alpha=0.25)
plt.show()
The output plot:

Numpy way to sort out a messy array for plotting

I have data of a plot on two arrays that are stored in unsorted way, so the plot jumps from one place to another discontinuously:
I have tried one example of finding the closest point in a 2D array:
import numpy as np
def distance(pt_1, pt_2):
pt_1 = np.array((pt_1[0], pt_1[1]))
pt_2 = np.array((pt_2[0], pt_2[1]))
return np.linalg.norm(pt_1-pt_2)
def closest_node(node, nodes):
nodes = np.asarray(nodes)
dist_2 = np.sum((nodes - node)**2, axis=1)
return np.argmin(dist_2)
a = []
for x in range(50000):
a.append((np.random.randint(0,1000),np.random.randint(0,1000)))
some_pt = (1, 2)
closest_node(some_pt, a)
Can I use it somehow to "clean" my data? (in the above code, a can be my data)
Exemplary data from my calculations is:
array([[ 2.08937872e+001, 1.99020033e+001, 2.28260611e+001,
6.27711094e+000, 3.30392288e+000, 1.30312878e+001,
8.80768833e+000, 1.31238275e+001, 1.57400130e+001,
5.00278061e+000, 1.70752624e+001, 1.79131456e+001,
1.50746185e+001, 2.50095731e+001, 2.15895974e+001,
1.23237801e+001, 1.14860312e+001, 1.44268222e+001,
6.37680265e+000, 7.81485403e+000],
[ -1.19702178e-001, -1.14050879e-001, -1.29711421e-001,
8.32977493e-001, 7.27437322e-001, 8.94389885e-001,
8.65931116e-001, -6.08199292e-002, -8.51922900e-002,
1.12333841e-001, -9.88131292e-324, 4.94065646e-324,
-9.88131292e-324, 4.94065646e-324, 4.94065646e-324,
0.00000000e+000, 0.00000000e+000, 0.00000000e+000,
-4.94065646e-324, 0.00000000e+000]])
After using radial_sort_line (of Joe Kington) I have received the following plot:

This is actually a problem that's tougher than you might think in general.
In your exact case, you might be able to get away with sorting by the y-values. It's hard to tell for sure from the plot.
Therefore, a better approach for somewhat circular shapes like this is to do a radial sort.
For example, let's generate some data somewhat similar to yours:
import numpy as np
import matplotlib.pyplot as plt
t = np.linspace(.2, 1.6 * np.pi)
x, y = np.cos(t), np.sin(t)
# Shuffle the points...
i = np.arange(t.size)
np.random.shuffle(i)
x, y = x[i], y[i]
fig, ax = plt.subplots()
ax.plot(x, y, color='lightblue')
ax.margins(0.05)
plt.show()
Okay, now let's try to undo that shuffle by using a radial sort. We'll use the centroid of the points as the center and calculate the angle to each point, then sort by that angle:
x0, y0 = x.mean(), y.mean()
angle = np.arctan2(y - y0, x - x0)
idx = angle.argsort()
x, y = x[idx], y[idx]
fig, ax = plt.subplots()
ax.plot(x, y, color='lightblue')
ax.margins(0.05)
plt.show()
Okay, pretty close! If we were working with a closed polygon, we'd be done.
However, we have one problem -- This closes the wrong gap. We'd rather have the angle start at the position of the largest gap in the line.
Therefore, we'll need to calculate the gap to each adjacent point on our new line and re-do the sort based on a new starting angle:
dx = np.diff(np.append(x, x[-1]))
dy = np.diff(np.append(y, y[-1]))
max_gap = np.abs(np.hypot(dx, dy)).argmax() + 1
x = np.append(x[max_gap:], x[:max_gap])
y = np.append(y[max_gap:], y[:max_gap])
Which results in:
As a complete, stand-alone example:
import numpy as np
import matplotlib.pyplot as plt
def main():
x, y = generate_data()
plot(x, y).set(title='Original data')
x, y = radial_sort_line(x, y)
plot(x, y).set(title='Sorted data')
plt.show()
def generate_data(num=50):
t = np.linspace(.2, 1.6 * np.pi, num)
x, y = np.cos(t), np.sin(t)
# Shuffle the points...
i = np.arange(t.size)
np.random.shuffle(i)
x, y = x[i], y[i]
return x, y
def radial_sort_line(x, y):
"""Sort unordered verts of an unclosed line by angle from their center."""
# Radial sort
x0, y0 = x.mean(), y.mean()
angle = np.arctan2(y - y0, x - x0)
idx = angle.argsort()
x, y = x[idx], y[idx]
# Split at opening in line
dx = np.diff(np.append(x, x[-1]))
dy = np.diff(np.append(y, y[-1]))
max_gap = np.abs(np.hypot(dx, dy)).argmax() + 1
x = np.append(x[max_gap:], x[:max_gap])
y = np.append(y[max_gap:], y[:max_gap])
return x, y
def plot(x, y):
fig, ax = plt.subplots()
ax.plot(x, y, color='lightblue')
ax.margins(0.05)
return ax
main()

Sorting the data base on their angle relative to the center as in #JoeKington 's solution might have problems with some parts of the data:
In [1]:
import scipy.spatial as ss
import matplotlib.pyplot as plt
import numpy as np
import re
%matplotlib inline
In [2]:
data=np.array([[ 2.08937872e+001, 1.99020033e+001, 2.28260611e+001,
6.27711094e+000, 3.30392288e+000, 1.30312878e+001,
8.80768833e+000, 1.31238275e+001, 1.57400130e+001,
5.00278061e+000, 1.70752624e+001, 1.79131456e+001,
1.50746185e+001, 2.50095731e+001, 2.15895974e+001,
1.23237801e+001, 1.14860312e+001, 1.44268222e+001,
6.37680265e+000, 7.81485403e+000],
[ -1.19702178e-001, -1.14050879e-001, -1.29711421e-001,
8.32977493e-001, 7.27437322e-001, 8.94389885e-001,
8.65931116e-001, -6.08199292e-002, -8.51922900e-002,
1.12333841e-001, -9.88131292e-324, 4.94065646e-324,
-9.88131292e-324, 4.94065646e-324, 4.94065646e-324,
0.00000000e+000, 0.00000000e+000, 0.00000000e+000,
-4.94065646e-324, 0.00000000e+000]])
In [3]:
plt.plot(data[0], data[1])
plt.title('Unsorted Data')
Out[3]:
<matplotlib.text.Text at 0x10a5c0550>
See x values between 15 and 20 are not sorted correctly.
In [10]:
#Calculate the angle in degrees of [0, 360]
sort_index = np.angle(np.dot((data.T-data.mean(1)), np.array([1.0, 1.0j])))
sort_index = np.where(sort_index>0, sort_index, sort_index+360)
#sorted the data by angle and plot them
sort_index = sort_index.argsort()
plt.plot(data[0][sort_index], data[1][sort_index])
plt.title('Data Sorted by angle relatively to the centroid')
plt.plot(data[0], data[1], 'r+')
Out[10]:
[<matplotlib.lines.Line2D at 0x10b009e10>]
We can sort the data based on a nearest neighbor approach, but since the x and y are of very different scale, the choice of distance metrics becomes an important issue. We will just try all the distance metrics available in scipy to get an idea:
In [7]:
def sort_dots(metrics, ax, start):
dist_m = ss.distance.squareform(ss.distance.pdist(data.T, metrics))
total_points = data.shape[1]
points_index = set(range(total_points))
sorted_index = []
target = start
ax.plot(data[0, target], data[1, target], 'o', markersize=16)
points_index.discard(target)
while len(points_index)>0:
candidate = list(points_index)
nneigbour = candidate[dist_m[target, candidate].argmin()]
points_index.discard(nneigbour)
points_index.discard(target)
#print points_index, target, nneigbour
sorted_index.append(target)
target = nneigbour
sorted_index.append(target)
ax.plot(data[0][sorted_index], data[1][sorted_index])
ax.set_title(metrics)
In [6]:
dmetrics = re.findall('pdist\(X\,\s+\'(.*)\'', ss.distance.pdist.__doc__)
In [8]:
f, axes = plt.subplots(4, 6, figsize=(16,10), sharex=True, sharey=True)
axes = axes.ravel()
for metrics, ax in zip(dmetrics, axes):
try:
sort_dots(metrics, ax, 5)
except:
ax.set_title(metrics + '(unsuitable)')
It looks like standardized euclidean and mahanalobis metrics give the best result. Note that we choose a starting point of the 6th data (index 5), it is the data point this the largest y value (use argmax to get the index, of course).
In [9]:
f, axes = plt.subplots(4, 6, figsize=(16,10), sharex=True, sharey=True)
axes = axes.ravel()
for metrics, ax in zip(dmetrics, axes):
try:
sort_dots(metrics, ax, 13)
except:
ax.set_title(metrics + '(unsuitable)')
This is what happens if you choose the starting point of max. x value (index 13). It appears that mahanalobis metrics is better than standardized euclidean as it is not affected by the starting point we choose.

If we do the assumption that the data are 2D and the x axis should be in an increasing fashion, then you could:
sort the x axis data, e.g. x_old and store the result in a different variable, e.g. x_new
for each element in the x_new find its index in the x_old array
re-order the elements in the y_axis array according to the indices that you got from previous step
I would do it with python list instead of numpy array due to list.index method been more easily manipulated than the numpy.where method.
E.g. (and assume that x_old and y_old are your previous numpy variables for x and y axis respectively)
import numpy as np
x_new_tmp = x_old.tolist()
y_new_tmp = y_old.tolist()
x_new = sorted(x_new_tmp)
y_new = [y_new_tmp[x_new_tmp.index(i)] for i in x_new]
Then you can plot x_new and y_new

Why is my scatterplot not rotating?

I am trying to create a 2d gaussian distribution and to rotate it by some degree.
import numpy as np
import matplotlib.pyplot as plt
x = np.random.normal(0, 15, 5000)
y = np.random.normal(0, 3, 5000)
X = np.array([x, y])
print X.shape
angle = 28
theta = np.pi * angle / 180
rotation = np.array([[np.cos(theta), -np.sin(theta)], [np.sin(theta), np.cos(theta)]])
X1 = np.dot(rotation, X)
print X1.shape
fig = plt.figure(figsize=(16, 8))
fig.add_subplot(2, 1, 1).scatter(x, y)
fig.add_subplot(2, 1, 2).scatter(X1[0], X1[:1])
plt.show()
What I expect to see here is a first scatterplot of gaussian and then the second one almost the same, but rotated by 28 degree. But instead I see this:

You just have an error in the way you index X1.
Currently, you plot X1[0] against X1[:1], but X1[:1] is the same as X1[0], as you are saying "all indices in the first dimension up to 1" (i.e. 0).
You just need to get rid of the colon - i.e. you need to plot X1[0] and X1[1].
This works:
fig.add_subplot(2, 1, 2).scatter(X1[0], X1[1])

Convolution of two rectangular pulses Python

I am attempting to find the convolution of two rectangular pulses.
No errors are being thrown - and I am getting a suitably shaped waveform output - however, the magnitude of my answer appears to be vastly too large, and I'm also unsure of how to fit a correct x/time axis to this convolution.
Additionally, the magnitude of the convolution seems to depend on the number of samples in the two pulses (essentially the sampling frequency) - which I would say is incorrect.
As I am attempting to model a continuous time signal, rather than discrete, I have set the sampling frequency very high.
Clearly I am doing something wrong - but what is it, and how do I correct it?
Thanks very much - and apologies if some of the code is not very "pythonic" (Recent Java convert)!
EDIT: Whilst attempting to evaluate this by hand, I have found that the time axis is too small by a factor of 2; again, I don't know why this would be
import numpy as np
import matplotlib.pyplot as plt
from sympy.functions.special import delta_functions as dlta
def stepFunction(t): #create pulses below from step-functions
retval = 0
if t == 0:
retval = 1
else:
retval = dlta.Heaviside(t)
return retval
def hT (t=0, start=0, dur=8, samples=1000):
time = np.linspace(start, start + dur, samples, True)
data = np.zeros(len(time))
hTArray = np.column_stack((time, data))
for row in range(len(hTArray)):
hTArray[row][1] = 2 * (stepFunction(hTArray[row][0] - 4) - stepFunction(hTArray[row][0] - 6))
return hTArray
def xT (t=0, start=0, dur=8, samples=1000):
time = np.linspace(start, start + dur, samples, True)
data = np.zeros(len(time))
hTArray = np.column_stack((time, data))
for row in range(len(hTArray)):
hTArray[row][1] = (stepFunction(hTArray[row][0]) - stepFunction(hTArray[row][0] - 4))
return hTArray
hTArray = hT() #populate two arrays with functions
xTArray = xT()
resCon = np.convolve(hTArray[:, 1], xTArray[:, 1]) #convolute signals/array data
Xaxis = np.linspace(hTArray[0][0], hTArray[len(hTArray) - 1][0],
len(resCon), endpoint=True) # create time axis, with same intervals as original functions
#Plot the functions & convolution
plt.plot(hTArray[:, 0], hTArray[:, 1], label=r'$x1(t)$')
plt.plot(xTArray[:, 0], xTArray[:, 1], label=r'$x2(t)$')
plt.plot(Xaxis, resCon)
plt.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3,
ncol=2, mode="expand", borderaxespad=0.)
ax = plt.gca()
ax.grid(True)
ax.spines['right'].set_color('none')
ax.spines['top'].set_color('none')
ax.xaxis.set_ticks_position('bottom')
ax.spines['bottom'].set_position(('data', 0))
ax.yaxis.set_ticks_position('left')
ax.spines['left'].set_position(('data', 0))
plt.show()

When you convolute discrete signals, you need to scale appropriately to keep the signal's energy (integral over |x(t)|²) constant:
import numpy as np
import matplotlib.pyplot as plt
n = 1000
t = np.linspace(0, 8, n)
T = t[1] - t[0] # sampling width
x1 = np.where(t<=4, 1, 0) # build input functions
x2 = np.where(np.logical_and(t>=4, t<=6), 2, 0)
y = np.convolve(x1, x2, mode='full') * T # scaled convolution
ty = np.linspace(0, 2*8, n*2-1) # double time interval
# plot results:
fg, ax = plt.subplots(1, 1)
ax.plot(t, x1, label="$x_1$")
ax.plot(t, x2, label="$x_2$")
ax.plot(ty, y, label="$x_1\\star x_2$")
ax.legend(loc='best')
ax.grid(True)
fg.canvas.draw()