Jupyter中Matplotlib图表可显示却无法保存的技术求助

在Jupyter Notebook中，Matplotlib生成的复合轴图表能够正常显示，但使用多种方法保存时均得到空白图片。

可复现代码 #

import numpy as np
import matplotlib.pyplot as plt
# %matplotlib inline
from scipy.optimize import fsolve
import ast
from mpl_toolkits.axisartist.parasite_axes import HostAxes, ParasiteAxes
import matplotlib.pyplot as plt

basicDict = {'Model_Name': 'TestPump_d258', 'max_pressure': ' 16 бар', 'liquid_pump_temp': ' -15C-110C', 'max_env_temp': ' 40С', 'engine_efficiency_class': ' IE2', 'Net_connection': ' 3~380V/50 Hz', 'rotation_frequency_nominal': ' 2900 1/min', 'Nominal_power_P2': ' P2 15 кВт', 'Nominal_currency': ' 27 А', 'Moisture_protection': ' IP54', 'Isolation_Class': ' F', 'Sucking_pipe_branch': ' DN100', 'Pushing_pipe_branch': ' DN100', 'building_length': ' 480 мм', 'frame': ' GG20', 'working_wheel': ' GG20', 'Shaft': ' 1.4021', 'Weight': ' 335 ', 'flow_rate_Q_m3_h': '[0.00, 11.77, 23.54, 35.31, 47.07, 58.84, 70.61, 82.38, 94.15, 96.04, 105.92, 117.69, 129.45, 141.22, 152.99, 164.76]', 'lift_H_m': '[22.96, 22.92, 22.93, 22.92, 22.80, 22.51, 22.02, 21.28, 20.27, 20.08, 18.98, 17.40, 15.55, 13.45, 11.14, 8.65]', 'Cavitation_NPSH_m': '[1.14, 1.17, 1.20, 1.24, 1.29, 1.37, 1.46, 1.59, 1.76, 1.79, 1.98, 2.26, 2.61, 3.05, 3.58, 4.23]', 'motor_power_P2_kW': '[3.57, 3.84, 4.21, 4.67, 5.17, 5.68, 6.20, 6.70, 7.17, 7.24, 7.60, 7.99, 8.35, 8.67, 8.98, 9.28]', 'efficiency': '[0.00, 4.00, 8.00, 12.00, 16.00, 20.00, 24.00, 28.00, 32.00, 36.00, 40.00, 44.00, 41.00, 38.00, 35.00, 32.00]', 'Scale_Table': '{"L":500,"B":430,"H":855,"C1":160,"B1":220,"A":160,"C2":120,"B2":180,"4-d 1":"4-18"}', 'pump_wiring_connection_figure': 'connection_figures/test_connection.png', 'pump_figure': 'pump_figures/Test_pump_image.png'}
Q = 100
H = 18
flow_rate_Q_m3_h = np.array(ast.literal_eval(basicDict["flow_rate_Q_m3_h"]))
lift_H_m = np.array(ast.literal_eval(basicDict["lift_H_m"]))
Cavitation_NPSH_m = np.array(ast.literal_eval(basicDict["Cavitation_NPSH_m"]))
motor_power_P2_kW = np.array(ast.literal_eval(basicDict["motor_power_P2_kW"]))
efficiency = np.array(ast.literal_eval(basicDict["efficiency"]))
def Formula(Q, H, EfficiancyValue):
    Hvx =(((Q/100*EfficiancyValue)/Q)**2)*H
    return Hvx
def calculate_Hvx_Line(Q, H, EfficiancyMax):
    calculatedPoints = []
    percents = range(0,EfficiancyMax)
    for each in percents:
        calculatedPoints.append(Formula(Q, H, each))
    return calculatedPoints
# Строим модели полиномиальной регрессии для данных
model5 = np.poly1d (np.polyfit (flow_rate_Q_m3_h, lift_H_m , 6))
# на графике будет сглаженая, построеная по модели версия
pumpcurve1 = []
for each in range(int(flow_rate_Q_m3_h.max())):
    pumpcurve1.append(model5(each))
# график по расчету
calculatedPoints = np.array(calculate_Hvx_Line(Q, H, 120))
# Ищем x точки соприкосновения двух моделей
def F(k):
    return Formula(Q, H, k)-model5(k)
Qnew = fsolve(F, 42)[0]
# y точки соприкосновения двух моделей
h = Formula(Q, H, Qnew)
# плотим графики и точку
fig = plt.figure()
host = fig.add_axes([1.2, 1.2, 1.2, 1.2], axes_class=HostAxes)
par1 = ParasiteAxes(host, sharex=host)
par2 = ParasiteAxes(host, sharex=host)
host.parasites.append(par1)
host.parasites.append(par2)
host.axis["right"].set_visible(False)
par1.axis["right"].set_visible(True)
par1.axis["right"].major_ticklabels.set_visible(True)
par1.axis["right"].label.set_visible(True)
# плотим графики и точку
p1, = host.plot([Qnew], [h], 'o', markersize=8, color = "r")
p1, = host.plot([Qnew], [h], 'o', markersize=3, color = "k")
p1, = host.plot(np.linspace(0, 120,120),calculatedPoints, label="calculated")
p1, = host.plot(np.linspace(0, len(pumpcurve1),len(pumpcurve1)),pumpcurve1, label="flow_rate_Q_m3_h")
max_x = (int(np.array(flow_rate_Q_m3_h).max())//10)*(10)+10
host.set_xlim([0,max_x])
host.plot([0, Qnew],[h, h], "k--")
host.text(2, h-1,
         fr"h (m) = {h:.2f}", color="k",fontsize=10)
max_y = (int(np.array(calculatedPoints).max())//5)*(5+1)
host.set_ylim([0,max_y])
host.plot([Qnew, Qnew],[0, h], "k--")
host.text(Qnew + 2, 0.5,
         fr"Q (m$^3$/s) = {Qnew:.2f}", color="k")
p2, = par2.plot(np.linspace(0, np.max(flow_rate_Q_m3_h),len(flow_rate_Q_m3_h)),efficiency, label="efficiency")
ef_max = (int(np.array(efficiency).max())//5)*(5+1)
par1.set_ylim(0, ef_max)
host.set_xlabel(r'Q (m$^3$/s)')
host.set_ylabel('h (m)')
par1.set_ylabel("efficiency (%)")
host.legend()
host.grid(True)
host.axis["left"].label.set_color(p1.get_color())
par1.axis["right"].label.set_color(p2.get_color())
plt.show()

尝试过的保存方法 #

• 方法1：

import cv2 as cv
from PIL import Image, ImageFilter
fig1 = plt.gcf()
plt.draw()
fig1.savefig('im.png')

width, height = fig.get_size_inches() * fig.get_dpi()
width , height = int(width), int(height)

• 方法2：

canvas = FigureCanvas(fig)
canvas.draw()
buf = canvas.buffer_rgba()
X = np.asarray(buf)
cv.imshow("ready",X.reshape(height, width,4))

• 方法3：

canvas = FigureCanvas(fig)
image = np.frombuffer(canvas.tostring_rgb(), dtype='uint8')
cv.imshow("ready",image.reshape(height, width, 3))

• 方法4：

def fig2img(fig):
    """Convert a Matplotlib figure to a PIL Image and return it"""
    import io
    buf = io.BytesIO()
    fig.savefig(buf)
    buf.seek(0)
    img = Image.open(buf)
    return img

im = np.uint8(fig2img(fig))
cv.imshow("ready",im)

cv.waitKey(00)

问题原因 #

1. Axes位置超出画布可视区域：创建HostAxes时使用的参数[1.2, 1.2, 1.2, 1.2]是[left, bottom, width, height]的归一化坐标（范围0-1），该参数把图表绘制在了画布默认区域之外，Jupyter显示时会自动调整画布范围，但保存时只会保存默认画布区域，导致空白。
2. plt.show()释放资源：plt.show()调用后，Matplotlib会释放figure的绘图资源，后续再保存figure会丢失内容。

有效解决方法 #

方法1：调整Axes位置并在plt.show()前保存 #

修改代码中创建HostAxes的部分，使用合理的归一化位置参数，并在plt.show()之前调用savefig：

# ... 保留原有代码 ...
# 调整host axes的位置参数到画布可视范围内
host = fig.add_axes([0.1, 0.1, 0.8, 0.8], axes_class=HostAxes)
# ... 保留原有绘图代码 ...

# 在plt.show()之前保存图片，bbox_inches='tight'确保所有元素被保存
fig.savefig('pump_chart.png', dpi=100, bbox_inches='tight')
plt.show()

方法2：重新绘制Figure后保存 #

如果必须在plt.show()之后保存，可以重新执行绘图逻辑（不调用plt.show()）再保存：

# 重新构建绘图逻辑，不调用plt.show()
fig = plt.figure()
host = fig.add_axes([0.1, 0.1, 0.8, 0.8], axes_class=HostAxes)
# ... 完整的绘图代码 ...

# 直接保存图片
fig.savefig('pump_chart.png', dpi=100, bbox_inches='tight')

方法3：使用Jupyter内置保存功能 #

在Jupyter中，图表显示后直接点击图表下方的保存按钮，选择PNG/SVG等格式，该方法会自动处理画布范围问题，适合快速保存。

内容的提问来源于stack exchange，提问作者Vitaly Skrinkovich