将平面拟合到二维阵列

首先,您需要以正确的方式表示数据。

你有两个论点 X1 和 X2 ,定义拓扑图像的坐标,以及一个目标值 Y ,定义每个点的高度。对于回归分析,您需要通过添加 X0 ,始终等于1。

然后需要将参数和目标展开为矩阵 [m*m x 3] 和 [m*m x 1] 分别地你想找到向量 theta ,其将描述期望的平面。为此,您可以使用 法线方程 :

为了演示该方法,我生成了一些拓扑曲面。您可以在图片上看到曲面、带有拟合平面的曲面以及相减后的曲面:

下面是代码:

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

m = 256 #size of the matrix

X1, X2 = np.mgrid[:m, :m]

fig = plt.figure()
ax = fig.add_subplot(3,1,1, projection='3d')
jet = plt.get_cmap('jet')

#generation of the surface
F = 3        
i = np.minimum(X1, m-X1-1)
j = np.minimum(X2, m-X2-1)
H = np.exp(-.5*(np.power(i, 2)  +  np.power(j, 2)   )/(F*F))
Y = np.real(  np.fft.ifft2   (H  *  np.fft.fft2(  np.random.randn(m, m))))
a = 0.0005; b = 0.0002; #parameters of the tilted plane
Y = Y + (a*X1 + b*X2); #adding the plane
Y = (Y - np.min(Y)) / (np.max(Y) - np.min(Y)) #data scaling

#plot the initial topological surface
ax.plot_surface(X1,X2,Y, rstride = 1, cstride = 1, cmap = jet, linewidth = 0)


#Regression
X = np.hstack(   ( np.reshape(X1, (m*m, 1)) , np.reshape(X2, (m*m, 1)) ) )
X = np.hstack(   ( np.ones((m*m, 1)) , X ))
YY = np.reshape(Y, (m*m, 1))

theta = np.dot(np.dot( np.linalg.pinv(np.dot(X.transpose(), X)), X.transpose()), YY)

plane = np.reshape(np.dot(X, theta), (m, m));

ax = fig.add_subplot(3,1,2, projection='3d')
ax.plot_surface(X1,X2,plane)
ax.plot_surface(X1,X2,Y, rstride = 1, cstride = 1, cmap = jet, linewidth = 0)


#Subtraction
Y_sub = Y - plane
ax = fig.add_subplot(3,1,3, projection='3d')
ax.plot_surface(X1,X2,Y_sub, rstride = 1, cstride = 1, cmap = jet, linewidth = 0)

plt.show()