在 matplotlib 中分层轮廓图和表面图

Question

I am struggling with layering and zorder in python. I am making a 3D plot using matplotlib with three relevant elements: A surface_plot of a planet, a surface_plot of rings around that planet, and a contourf image that shows the planet's shadow cast onto the rings.

I want the graphics to display exactly how this scenario would look in real life, with the rings going around the planet and the shadow residing across the rings in the appropriate spot. If the shadow is behind the planet for a given POV, I want the shadow to be blocked by the planet, and vice versa if the shadow is in front of the planet for a given POV.

To be clear, this is ONLY a layering issue. I have the planet, rings, and shadow all plotting correctly. However, the shadow will not ever display in front of the planet. It acts as though the planet is "blocking" the shadow, even though the planet is supposed to be underneath the shadow in terms of layering.

I have tried every single thing I can think of in terms of zorder and rearranging which order the various plot elements are called to be drawn. The rings DO correctly display in front of the planet, but the shadow will not.

My actual code is very long. here are the relevant parts:

Plot setup:

def orthographic_proj(zfront, zback):
    a = (zfront+zback)/(zfront-zback)
    b = -2*(zfront*zback)/(zfront-zback)
    return np.array([[1,0,0,0],
                        [0,1,0,0],
                        [0,0,a,b],
                        [0,0,0,zback]])

def setup_saturn_plot(ax3, elev, azim, drawz, drawxy,view):
    #ax3.set_aspect('equal','box')
    ax3.view_init(elev=elev, azim=azim)
    if(view=="top" or view == "Top" or view == "TOP"):
        ax3.dist = 5.5
    if(view=="star" or view == "Star" or view == "STAR"):
        ax3.dist = 5.0 #4.5 is best value
    proj3d.persp_transformation = orthographic_proj

    # hide grid and background
    ax3.w_xaxis.set_pane_color((1.0, 1.0, 1.0, 1.0))
    ax3.w_yaxis.set_pane_color((1.0, 1.0, 1.0, 1.0))
    ax3.w_zaxis.set_pane_color((1.0, 1.0, 1.0, 1.0))
    ax3.grid(False)

    # hide z axis in orthographic top view, xy axes in star view
    if (drawz == False):
        ax3.w_zaxis.line.set_lw(0.)
        ax3.set_zticks([])

    if (drawz == True):
        ax3.set_zlabel('Z (1000 km)',fontsize=12)

    if (drawxy == False):
        ax3.w_xaxis.line.set_lw(0.)
        ax3.set_xticks([])
        ax3.w_yaxis.line.set_lw(0.)
        ax3.set_yticks([])

    if (drawxy == True):
        ax3.set_xlabel('X (1000 km)',fontsize=12)
        ax3.set_ylabel('Y (1000 km)',fontsize=12)

Planet:

def draw_saturn(ax3, elev, azim):
    # Saturn dimensions
    radius = 60268. / 1000.
    radius_pole = 54364. / 1000.

    # draw Saturn
    phi, theta = np.mgrid[0.0:np.pi:100j, 0.0:2.0*np.pi:100j]
    x = radius*np.sin(phi)*np.cos(theta)
    y = radius*np.sin(phi)*np.sin(theta)
    z = radius_pole*np.cos(phi)

    line3 = ax3.plot_surface(x, y, z, color="w", edgecolor='b', rstride = 8, cstride=5, shade=False, lw=0.25)
    #line3 = ax3.plot_wireframe(x, y, z, color="w", edgecolor='b', rstride = 5, cstride=5, lw=0.25)

    ax3.tick_params(labelsize=10)

rings:

def draw_rings(ax3, elev, azim, draw_mode):
    # Saturn dimensions
    radius = 60268. / 1000.

    # Saturn rings
    dringmin = 1.110 * radius 
    dringmax = 1.236 * radius 
    cringmin = 1.239 * radius 
    titanringlet = 1.292 * radius 
    maxwellgap = 1.452 * radius 
    cringmax = 1.526 * radius 
    bringmin = 1.526 * radius 
    bringmax = 1.950 * radius 
    aringmin = 2.030 * radius 
    enckegap = 2.214 * radius 
    keelergap = 2.265 * radius 
    aringmax = 2.270 * radius 
    fringmin = 2.320 * radius 
    gringmin = 2.754 * radius 
    gringmax = 2.874 * radius 
    eringmin = 2.987 * radius 
    eringmax = 7.964 * radius 

    if (draw_mode == 'back'):
        offset = -azim*np.pi/180. - 0.5*np.pi
    if (draw_mode == 'front'):
        offset = -azim*np.pi/180. + 0.5*np.pi

    rad, theta = np.mgrid[dringmin:dringmax:4j, 0.0-offset:1.0*np.pi-offset:100j]
    x = rad * np.cos(theta)
    y = rad * np.sin(theta)
    z = 0. * rad
    line1 = ax3.plot_surface(x, y, z, color="w", edgecolor='b', rstride = 8, cstride=25, shade=False, lw=0.25,alpha=0.)

    rad, theta = np.mgrid[cringmin:cringmax:4j, 0.0-offset:1.0*np.pi-offset:100j]
    x = rad * np.cos(theta)
    y = rad * np.sin(theta)
    z = 0. * rad
    line2 = ax3.plot_surface(x, y, z, color="w", edgecolor='b', rstride = 8, cstride=25, shade=False, lw=0.25,alpha=0.)

    rad, theta = np.mgrid[bringmin:bringmax:4j, 0.0-offset:1.0*np.pi-offset:100j]
    x = rad * np.cos(theta)
    y = rad * np.sin(theta)
    z = 0. * rad
    line3 = ax3.plot_surface(x, y, z, color="w", edgecolor='b', rstride = 8, cstride=25, shade=False, lw=0.25,alpha=0.)

    rad, theta = np.mgrid[aringmin:aringmax:4j, 0.0-offset:1.0*np.pi-offset:100j]
    x = rad * np.cos(theta)
    y = rad * np.sin(theta)
    z = 0. * rad
    line4 = ax3.plot_surface(x, y, z, color="w", edgecolor='b', rstride = 8, cstride=25, shade=False, lw=0.25,alpha=0.)

    rad, theta = np.mgrid[fringmin:1.005*fringmin:2j, 0.0-offset:1.0*np.pi-offset:100j]
    x = rad * np.cos(theta)
    y = rad * np.sin(theta)
    z = 0. * rad
    line7 = ax3.plot_surface(x, y, z, color="w", edgecolor='b', rstride = 8, cstride=25, shade=False, lw=0.1,alpha=0.)

Shadow:

def draw_shadowboundary(ax3, sundir):
    sqrt = np.sqrt

    #azimuthal angle between x direction and direction of sun
    alpha = np.arctan2(sundir[1],sundir[0])
    #adjustments to keep -pi/2 < alpha < pi/2
    alphaadj = 0.*np.pi/180.
    if (alpha<0.):
        alpha += 2.*np.pi
    if ((alpha >= np.pi/2.) & (alpha <= np.pi)):
        alpha += np.pi
        alphaadj = np.pi
    if ((alpha > np.pi) & (alpha <= 3.*np.pi/2.)):
        alpha -= np.pi
        alphaadj = np.pi
    if (alpha>3.*np.pi/2.):
        alpha-=2*np.pi 

    #azimuthal angle between x direction and northern summer -- found using VIMS_2005_14_OMICET and VIMS_2017_053_ALPORI to define eq. of plane of Sun's annual path in chosen coordinate system: -0.193318*x + 0.1963755*y + 0.5471502*z = 0
    beta = 44.5505*np.pi/180.
    #Saturn's obliquity -- from NASA fact sheet
    psi = 26.73*np.pi/180.
    #Saturn's oblateness -- from NASA fact sheet
    obl = 0.09796
    #helpful definitions for optimization
    cpsic = np.cos(psi*np.cos(alpha+beta))
    spsic = np.sin(psi*np.cos(alpha+beta))
    calpha = np.cos(alpha)
    salpha = np.sin(alpha)
    #Saturn's projected shorter planetary axis as seen by the sun & ring inner edge
    req = 60268. / 1000.    
    b = req*sqrt((1.-obl)*(1.-obl)*cpsic*cpsic + spsic*spsic)
    ringstart = 1.239 * req
    ringend = 2.270 * req
    #shadow boundary of Saturn's rings -- can approximate using a=inf and cancelling terms
    a = 9.582*1.496*10.**5
    shadowline = lambda x,y : (1/a)*sqrt((req*salpha*(-a+x*calpha*cpsic+y*salpha)*(y*calpha-x*cpsic*salpha)/sqrt((y*calpha-x*cpsic*salpha)**2 + (x*spsic)**2) + calpha*(a*cpsic*(x*calpha*cpsic+y*salpha) + b*x*(a-x*calpha*cpsic-y*salpha)*spsic*spsic/sqrt((y*calpha-x*cpsic*salpha)**2 + (x*spsic)**2)))**2 + (req*calpha*(a-x*calpha*cpsic-y*salpha)*(y*calpha-x*cpsic*salpha)/sqrt((y*calpha-x*cpsic*salpha)**2 + (x*spsic)**2) + salpha*(a*cpsic*(x*calpha*cpsic+y*salpha)+b*x*(a-x*calpha*cpsic-y*salpha)*spsic*spsic/sqrt((y*calpha-x*cpsic*salpha)**2 + (x*spsic)**2)))**2)                                                                                                      
    #azimuthal radius & antisolar angle for inequalities
    radius = lambda x,y : np.sqrt(x**2+y**2)
    anti = lambda x,y : abs(np.arctan2(y,x)-(alpha-alphaadj))

    #properties of shadow
    samples=1200
    d = np.linspace(-3*req,3*req,samples)
    x,y = np.meshgrid(d,d)
    #z = ((radius(x,y)<=shadowline(x,y)) & (ringstart<=radius(x,y)) & (np.pi/2<=anti(x,y)) & (anti(x,y)<=3.*np.pi/2)).astype(int)
    z = ((radius(x,y)<=shadowline(x,y)) & (ringstart<=radius(x,y)) & (radius(x,y)<=ringend) & (np.pi/2<=anti(x,y)) & (anti(x,y)<=3.*np.pi/2)).astype(int)
    cmap = matplotlib.colors.ListedColormap(["k","k"])
    #add shadow to plot
    ax3.contourf(x,y,z, [0.5,1.50001], cmap=cmap,alpha=0.5)

Combine graphics:

import matplotlib
import numpy

from math import *

import matplotlib.pyplot as plt
import numpy as np
from mpl_toolkits.mplot3d import Axes3D # <--- This is important for 3d plotting 

from mpl_toolkits.mplot3d import proj3d

def plot_results(phi, theta, sundir=[0.5, 0.5]):
    #plot_names.append("occultation_track_" + starname)
    fig2 = plt.figure(figsize=(9,9))
    ax3 = fig2.add_subplot(111, projection='3d')
    setup_saturn_plot(ax3, phi, theta, False, False, "star")
    draw_saturn(ax3, phi, theta)
    draw_rings(ax3, phi, theta, 'back')
    draw_rings(ax3, phi, theta, 'front')
    draw_shadowboundary(ax3,sundir)
    ax3.set_xlim([-200, 200]) 
    ax3.set_ylim([-200, 200])
    ax3.set_zlim([-200, 200])


plot_results(phi=40, theta=50, sundir = (30,60))

The code produces an image like this:

The grey shadow is supposed to be residing on the rings in front of the planet. However, it won't display in front of the planet, so only the little sliver of shadow to the right of the planet is actually appearing. The shadow displays correctly in all scenarios except when it needs to go in front the planet.

Any fixes for this?

Answer 1

I'm working on getting my head around this code at the moment, but in the meantime, at least so far, this seams to be a known issue with matplotlib3d.

As @TheImportanceOfBeingErnest pointed out a long time ago, this issue appears in the mpl3d faq

My 3D plot doesn't look right at certain viewing angles

This is probably the most commonly reported issue with mplot3d. The problem is that – from some viewing angles – a 3D object would appear in front of another object, even though it is physically behind it. This can result in plots that do not look “physically correct.”

Unfortunately, while some work is being done to reduce the occurrence of this artifact, it is currently an intractable problem, and can not be fully solved until matplotlib supports 3D graphics rendering at its core.

The problem occurs due to the reduction of 3D data down to 2D + z-order scalar. A single value represents the 3rd dimension for all parts of 3D objects in a collection. Therefore, when the bounding boxes of two collections intersect, it becomes possible for this artifact to occur. Furthermore, the intersection of two 3D objects (such as polygons or patches) can not be rendered properly in matplotlib's 2D rendering engine.

This problem will likely not be solved until OpenGL support is added to all of the backends (patches are greatly welcomed). Until then, if you need complex 3D scenes, we recommend using MayaVi.