ax.plot(Xtrue[0],Xtrue[1],Xtrue[2],'r.')
plt.axis('equal')
#5.2.2 3Dの点群からカメラ行列を計算する
corr = corr[:,0] #
ndx3D = np.where(corr>=0)[0]
ndx2D = corr[ndx3D]
#見える点を選び同次座標に変換
x = points2D[0][:,ndx2D] #
x = np.vstack( (x,np.ones(x.shape[1])) )
X = points3D[:,ndx3D]
X = np.vstack( (X,np.ones(X.shape[1])) )
#Pを指定する
Pest = camera.Camera(sfm.compute_P(x,X))
#比較する
print Pest.P / Pest.P[2,3]
print P[0].P / P[0].P[2,3]
xest = Pest.project(X)
#描画する
plt.figure()
plt.imshow(im1)
plt.plot(x[0],x[1],'bo')
plt.plot(xest[0],xest[1],'r.')
plt.axis('off')
plt.show
# +
# %matplotlib inline
corr1 = corr[:, 0] #第一視点のインデックスを取り出す。
ndx3D = np.where(corr1 >= 0)[0]
ndx2D = corr1[ndx3D]
#見える点を選び、同次座標に変換
x = points2D[0][:, ndx2D]
x = homography.make_homog(x)
X = points3D[:, ndx3D]
X = homography.make_homog(X)
# Pを推定する
Pest = camera.Camera(sfm.compute_P(x, X))
#比較する
print(Pest.P / Pest.P[2, 3])
print(P[0].P / P[0].P[2, 3])
xest = Pest.project(X)
#描写する。
ip.show_img(im1)
plt.plot(x[0], x[1], 'bo')
plt.plot(xest[0], xest[1], 'r.')
plt.axis('off')
plt.show()
# -
import pandas as pd
import numpy as np
points2D = [np.loadtxt('2D/00' + str(i + 1) + '.corners').T for i in range(3)]
points3D = np.loadtxt('3D/p3d').T
corr = np.genfromtxt('2D/nview-corners', dtype='int', missing_values='*')
corr = corr[:, 0]
ndx3D = np.where(corr >= 0)[0] # missing values are -1
ndx2D = corr[ndx3D]
x = points2D[0][:, ndx2D] # view 1
x = np.vstack((x, np.ones(x.shape[1])))
X = points3D[:, ndx3D]
X = np.vstack((X, np.ones(X.shape[1])))
# estimate P
print sfm.compute_P(x, X)
corr = np.genfromtxt('2D/nview-corners', dtype='int', missing_values='*')
# index for points in first two views
ndx = (corr[:, 0] >= 0) & (corr[:, 1] >= 0)
# get coordinates and make homogeneous
x1 = points2D[0][:, corr[ndx, 0]]
x1 = np.vstack((x1, np.ones(x1.shape[1])))
x2 = points2D[1][:, corr[ndx, 1]]
x2 = np.vstack((x2, np.ones(x2.shape[1])))
# compute F
print x1
print x2
F = sfm.compute_fundamental(x1, x2)
print 'F', F