用 matplotlib 表示体素

Representing voxels with matplotlib

在 Python 中,给定一个包含 0 或 1 的 N_1 x N_2 x N_3 矩阵,我会寻找一种方法将 3D 数据显示为具有体积像素的 N_1 x N_2 x N_3 体积(体素)在 1s 的位置。

例如,如果 1s 的坐标是 [[1, 1, 1], [4, 1, 2], [3, 4, 1]],则所需的输出将如下所示

matplotlib 的mplot3D 模块似乎有可能实现这一点,但我还没有找到这种情节的任何例子。有人知道解决此问题的简单解决方案吗?

非常感谢您的帮助。

一个。使用 voxels

从 matplotlib 2.1 开始,有一个 Axes3D.voxels 函数可用,它几乎可以满足此处的要求。然而,它不是很容易定制成不同的尺寸、位置或颜色。

from mpl_toolkits.mplot3d import Axes3D
import numpy as np
import matplotlib.pyplot as plt
N1 = 10
N2 = 10
N3 = 10
ma = np.random.choice([0,1], size=(N1,N2,N3), p=[0.99, 0.01])

fig = plt.figure()
ax = fig.gca(projection='3d')
ax.set_aspect('equal')

ax.voxels(ma, edgecolor="k")

plt.show()

要将体素放置在不同的位置,请参阅

乙。使用 Poly3DCollection

手动创建体素可能会使过程更加透明,并允许对体素的大小、位置和颜色进行任何类型的自定义。另一个优点是,在这里我们创建了一个 Poly3DCollection 而不是许多,使这个解决方案比内置 voxels.

更快
from mpl_toolkits.mplot3d import Axes3D
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d.art3d import Poly3DCollection

def cuboid_data(o, size=(1,1,1)):
    X = [[[0, 1, 0], [0, 0, 0], [1, 0, 0], [1, 1, 0]],
         [[0, 0, 0], [0, 0, 1], [1, 0, 1], [1, 0, 0]],
         [[1, 0, 1], [1, 0, 0], [1, 1, 0], [1, 1, 1]],
         [[0, 0, 1], [0, 0, 0], [0, 1, 0], [0, 1, 1]],
         [[0, 1, 0], [0, 1, 1], [1, 1, 1], [1, 1, 0]],
         [[0, 1, 1], [0, 0, 1], [1, 0, 1], [1, 1, 1]]]
    X = np.array(X).astype(float)
    for i in range(3):
        X[:,:,i] *= size[i]
    X += np.array(o)
    return X

def plotCubeAt(positions,sizes=None,colors=None, **kwargs):
    if not isinstance(colors,(list,np.ndarray)): colors=["C0"]*len(positions)
    if not isinstance(sizes,(list,np.ndarray)): sizes=[(1,1,1)]*len(positions)
    g = []
    for p,s,c in zip(positions,sizes,colors):
        g.append( cuboid_data(p, size=s) )
    return Poly3DCollection(np.concatenate(g),  
                            facecolors=np.repeat(colors,6, axis=0), **kwargs)

N1 = 10
N2 = 10
N3 = 10
ma = np.random.choice([0,1], size=(N1,N2,N3), p=[0.99, 0.01])
x,y,z = np.indices((N1,N2,N3))-.5
positions = np.c_[x[ma==1],y[ma==1],z[ma==1]]
colors= np.random.rand(len(positions),3)

fig = plt.figure()
ax = fig.gca(projection='3d')
ax.set_aspect('equal')

pc = plotCubeAt(positions, colors=colors,edgecolor="k")
ax.add_collection3d(pc)

ax.set_xlim([0,10])
ax.set_ylim([0,10])
ax.set_zlim([0,10])
#plotMatrix(ax, ma)
#ax.voxels(ma, edgecolor="k")

plt.show()

C。使用 plot_surface

正在改编来自 (which is partly based on ), one can easily plot cuboids as surface plots 的代码。

然后可以遍历输入数组并找到 1 在对应于数组索引的位置绘制一个长方体。

这里的优点是您可以在表面上获得漂亮的阴影,增加 3D 效果。缺点可能是立方体在某些情况下可能不会表现出物理行为,例如它们可能会在某些视角重叠。

from mpl_toolkits.mplot3d import Axes3D
import numpy as np
import matplotlib.pyplot as plt

def cuboid_data(pos, size=(1,1,1)):
    # code taken from
    # 
    # suppose axis direction: x: to left; y: to inside; z: to upper
    # get the (left, outside, bottom) point
    o = [a - b / 2 for a, b in zip(pos, size)]
    # get the length, width, and height
    l, w, h = size
    x = [[o[0], o[0] + l, o[0] + l, o[0], o[0]],  
         [o[0], o[0] + l, o[0] + l, o[0], o[0]],  
         [o[0], o[0] + l, o[0] + l, o[0], o[0]],  
         [o[0], o[0] + l, o[0] + l, o[0], o[0]]]  
    y = [[o[1], o[1], o[1] + w, o[1] + w, o[1]],  
         [o[1], o[1], o[1] + w, o[1] + w, o[1]],  
         [o[1], o[1], o[1], o[1], o[1]],          
         [o[1] + w, o[1] + w, o[1] + w, o[1] + w, o[1] + w]]   
    z = [[o[2], o[2], o[2], o[2], o[2]],                       
         [o[2] + h, o[2] + h, o[2] + h, o[2] + h, o[2] + h],   
         [o[2], o[2], o[2] + h, o[2] + h, o[2]],               
         [o[2], o[2], o[2] + h, o[2] + h, o[2]]]               
    return np.array(x), np.array(y), np.array(z)

def plotCubeAt(pos=(0,0,0),ax=None):
    # Plotting a cube element at position pos
    if ax !=None:
        X, Y, Z = cuboid_data( pos )
        ax.plot_surface(X, Y, Z, color='b', rstride=1, cstride=1, alpha=1)

def plotMatrix(ax, matrix):
    # plot a Matrix 
    for i in range(matrix.shape[0]):
        for j in range(matrix.shape[1]):
            for k in range(matrix.shape[2]):
                if matrix[i,j,k] == 1:
                    # to have the 
                    plotCubeAt(pos=(i-0.5,j-0.5,k-0.5), ax=ax)            

N1 = 10
N2 = 10
N3 = 10
ma = np.random.choice([0,1], size=(N1,N2,N3), p=[0.99, 0.01])

fig = plt.figure()
ax = fig.gca(projection='3d')
ax.set_aspect('equal')

plotMatrix(ax, ma)

plt.show()

即将推出的matplotlib 2.1版确实有一个功能examples for 3D voxels

如果你使用 anaconda,你可以通过 conda-forge 频道安装它。

conda install -c conda-forge matplotlib