7

I amtrying to use scipy.spatial (from scipy.spatial import ConvexHull)to draw convex hull of series of points.

import pylab as pl
from scipy.spatial import ConvexHull

pl.figure()  
pl.hold(True)  

points = np.concatenate((x, y), axis=1)

hull = ConvexHull(points)

pl.plot(points[:,0], points[:,1], 'ro')

for simplex in hull.simplices:
    pl.plot(points[simplex,0], points[simplex,1], 'dk--')

The problem is I don't correctly understand what is hull.simplices, I want to find the indices of points which are on the facet of convexhull so I can use these indices to get the point from x and y

Am1rr3zA
  • 7,115
  • 18
  • 83
  • 125

1 Answers1

17

In the 2-D case, the simplices attribute of the ConvexHull object holds the pairs of indices of the points that make up the line segments of the convex hull. One way to get just the indices is to get the unique elements of the flattened simplices array. But note that the points will not be in an order that follows the convex hull around the set. (In scipy 0.13.0 and later, you can use the vertices attribute to get the indices; see below.)

For example,

import numpy as np
from scipy.spatial import ConvexHull
import matplotlib.pyplot as plt


# Generate some random points for the demo.
np.random.seed(4321)
pts = 0.1 + 0.8*np.random.rand(15, 2)

ch = ConvexHull(pts)

# hull_indices = ch.vertices   # This will work in the scipy 0.13
hull_indices = np.unique(ch.simplices.flat)
hull_pts = pts[hull_indices, :]

plt.plot(pts[:, 0], pts[:, 1], 'ko', markersize=10)
plt.plot(hull_pts[:, 0], hull_pts[:, 1], 'ro', alpha=.25, markersize=20)
plt.xlim(0, 1)
plt.ylim(0, 1)
plt.show()

This generates:

Plot of points and convex hull

The vertices attribute was added in scipy 0.13.0:

import numpy as np
from scipy.spatial import ConvexHull
import matplotlib.pyplot as plt


# Generate some random points for the demo.
np.random.seed(4321)
pts = 0.1 + 0.8*np.random.rand(15, 2)

ch = ConvexHull(pts)

# Get the indices of the hull points.
hull_indices = ch.vertices

# These are the actual points.
hull_pts = pts[hull_indices, :]

plt.plot(pts[:, 0], pts[:, 1], 'ko', markersize=10)
plt.fill(hull_pts[:,0], hull_pts[:,1], fill=False, edgecolor='b')
plt.xlim(0, 1)
plt.ylim(0, 1)
plt.show()

convex hull example

Warren Weckesser
  • 110,654
  • 19
  • 194
  • 214
  • I saw vertices before but my problem is that when I use Vertices I got this error "AttributeError: 'ConvexHull' object has no attribute 'vertices'" – Am1rr3zA Aug 11 '13 at 19:01
  • Ah, sorry. I was using the latest version of scipy from github. The `vertices` attribute was added just a few months ago (https://github.com/scipy/scipy/commit/3ff9d5fb2d2b9de660ee39a62a19ff186fc1618a). – Warren Weckesser Aug 11 '13 at 21:44
  • thanks for your solution but I have one big problem with your solution if you want to draw convexhull based on their index (pl.plot(points[hull_indices,0], points[hull_indices,1], 'b--')) the shape is not correct, I think that's because your solution didn't save the order of index – Am1rr3zA Aug 12 '13 at 20:53
  • That's what I meant when I said "the points will not be in an order that follows the convex hull around the set" (and it's why I used special markers instead of drawing line segments). The docstring for ConvexHull shows an example of plotting the shape: http://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.ConvexHull.html#scipy.spatial.ConvexHull – Warren Weckesser Aug 12 '13 at 21:08