使用 Python 的 matplotlib 3D API 的轮廓问题

Question

我正在尝试做一些类似于文档中的 3D 示例的事情，但使用点云而不是光滑表面。该示例将 2D 轮廓投影到三个坐标平面中的每一个平面上。这表明我能够在xy平面上做到这一点。当我尝试在其他两个平面上做同样的事情时，我得到一个奇怪的轮廓塌陷成一条细条，或者是我在 matplotlib 内部无法企及的例外。Traceback (most recent call last):  File ".../matplotlib/backends/backend_qt5.py", line 519, in _draw_idle    self.draw()  File ".../matplotlib/backends/backend_agg.py", line 433, in draw    self.figure.draw(self.renderer)  File ".../matplotlib/artist.py", line 55, in draw_wrapper    return draw(artist, renderer, *args, **kwargs)  File ".../matplotlib/figure.py", line 1475, in draw    renderer, self, artists, self.suppressComposite)  File ".../matplotlib/image.py", line 141, in _draw_list_compositing_images    a.draw(renderer)  File ".../mpl_toolkits/mplot3d/axes3d.py", line 281, in draw    reverse=True)):  File ".../mpl_toolkits/mplot3d/axes3d.py", line 280, in <lambda>    key=lambda col: col.do_3d_projection(renderer),  File ".../mpl_toolkits/mplot3d/art3d.py", line 226, in do_3d_projection    self._segments3d]  File ".../mpl_toolkits/mplot3d/art3d.py", line 225, in <listcomp>    proj3d.proj_trans_points(points, renderer.M) for points in  File ".../mpl_toolkits/mplot3d/proj3d.py", line 188, in proj_trans_points    xs, ys, zs = zip(*points)ValueError: not enough values to unpack (expected 3, got 0)下面是该问题的一个示例。此版本有效。取消注释函数中的一个或两个调用，以查看奇怪的轮廓，或者更有可能是异常。ax.contour()plot()这是 matplotlib 2.2.2 和 3.1.1 的。感谢您提供的任何帮助，以获得所有三个平面上的轮廓，例如演示。

Answer 1

一位matplotlib开发人员指出，重新采样是错误的。修复后，这是更正后的情节。

对于边上看到数据的坐标平面，例如本例中的 yz 平面，等高线看起来可能有点不稳定。这是意料之中的，因为数据可以接近多值。xz平面轮廓看起来也很粗糙。我怀疑这两个问题都可以通过单独三角化和轮廓化每个平面来改善，而不是像这里所做的那样偏爱 xy 平面。

下面是固定的测试脚本。唯一重要的更改是 和 。plot()resample()

import math

import sys

import matplotlib as mpl

import matplotlib.pyplot as plt

import mpl_toolkits.mplot3d

import numpy as np

import scipy.spatial

#np.random.seed(4)

numPts = 1000                                       # number of points in cloud

numGrid = 120      # number of points in meshgrid used in resample for contours

scatter = False                         # adds scatter plot to show point cloud

def main():

    (pts, f) = mkData()                                # create the point cloud

    tris = mkTris(pts)                                         # triangulate it

    plot(pts, tris, f)                                                # plot it

def plot(pts, tris, f):

    fig = plt.figure()

    ax = fig.add_subplot(111, projection="3d")

    cmap = plt.get_cmap("coolwarm")

    collec = ax.plot_trisurf(tris, pts[:, 2], cmap=cmap)

    colors = np.mean(f[tris.triangles], axis=1)

    collec.set_array(colors)

    collec.autoscale()

    (xr, yr, zr, fr, xMin, xMax, yMin, yMax, zMin, zMax) = resample(ax, tris,

       pts, f)

    ax.set_xlim(xMin, xMax)      # default always includes zero for some reason

    ax.set_ylim(yMin, yMax)

    ax.set_zlim(zMin, zMax)

    ax.contour(xr, yr, fr, 10, zdir="z", cmap=cmap, offset=zMin)

    ax.contour(fr, yr, zr, 10, zdir="x", cmap=cmap, offset=xMin)

    ax.contour(xr, fr, zr, 10, zdir="y", cmap=cmap, offset=yMax)

    if scatter:

       ax.scatter(pts[:, 0], pts[:, 1], pts[:, 2], alpha=0.1)

    ax.set_xlabel("x")

    ax.set_ylabel("y")

    ax.set_zlabel("z")

    fig.colorbar(collec, shrink=0.5, aspect=5)

    plt.show()

def mkData():

    """

    Create a random point cloud near a random plane, and define a function on

    the plane for colors and contours.

    """

    offset = 1                   # generate points near a unit square, xy-plane

    pts = 2 * np.random.rand(numPts, 3) - 1

    pts[:, 2] = offset * (2 * np.random.rand(numPts) - 1)

    x = 2 * np.ravel(pts[:, 0])

    y = 2 * np.ravel(pts[:, 1])

    f = x * np.exp(-x**2 - y**2)      # just some function for colors, contours

    width = 100                       # scale unit square to a larger rectangle

    height = 20

    pts[:, 0] *= width

    pts[:, 1] *= height

    (e1, e2, e3) =[2 * np.pi * np.random.rand() for _ in range(3)]

    (c1, s1) = (math.cos(e1), math.sin(e1))           # rotate scaled rectangle

    (c2, s2) = (math.cos(e2), math.sin(e2))

    (c3, s3) = (math.cos(e3), math.sin(e3))

    Ta2b = np.array((                     # do not have scipy.spatial.transform

        [               c1  *c2,       s2,              -s1 * c2],

        [s1 * s3 - c1 * s2 * c3,  c2 * c3, c1 *s3 + s1 * s2 * c3],

        [s1 * c3 + c1 * s2 * s3, -c2 * s3, c1 *c3 - s1 * s2 * s3]))

    pts = (Ta2b @ pts.T).T

    dist = 500                                 # translate away from the origin

    Ra2bNb = dist * (2 * np.random.rand(3, 1) - 1)

    pts += Ra2bNb.T

    return (pts, f)

def mkTris(pts):

    "Triangulate the point cloud."

    u = np.ravel(pts[:, 0])

    v = np.ravel(pts[:, 1])

    try:

        return mpl.tri.Triangulation(u, v)

    except (ValueError, RuntimeError) as ex:

        sys.exit(f"Unable to compute triangulation: {ex}.")

def resample(ax, tris, pts, f):

    "Resample the triangulation onto a regular grid for contours."

    (xMin, xMax) = ax.get_xlim()

    (yMin, yMax) = ax.get_ylim()

    (zMin, zMax) = ax.get_zlim()

    x = np.linspace(xMin, xMax, numGrid)

    y = np.linspace(yMin, yMax, numGrid)

    (xm, ym)=np.meshgrid(x, y)

    fInterp = mpl.tri.CubicTriInterpolator(tris, f)

    fm = fInterp(xm, ym)

    zInterp = mpl.tri.CubicTriInterpolator(tris, pts[:,2])

    zm = zInterp(xm, ym)

    return (xm, ym, zm, fm, xMin, xMax, yMin, yMax, zMin, zMax)

if __name__ == "__main__":

    main()