def triangulatation_self(points1n, points2n):
# achieved self
# points1 = img_pnts[0]
# points2 = img_pnts[1]
# one_vec = np.ones((1, points1.shape[1]))
# K_inv = np.linalg.inv(K)
# pnts1 = np.vstack((points1, one_vec))
# pnts2 = np.vstack((points2, one_vec))
# points1n = np.dot(K_inv, pnts1) # 3 x n
# points2n = np.dot(K_inv, pnts2) # 3 x n
E = structure.compute_essential_normalized(points1n, points2n)
print('Computed essential matrix:', (-E / E[0][1]))
# Given we are at camera 1, calculate the parameters for camera 2
# Using the essential matrix returns 4 possible camera paramters
P1 = np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])
P2s = structure.compute_P_from_essential(E)
ind = -1
for i, P2 in enumerate(P2s):
# Find the correct camera parameters
d1 = structure.reconstruct_one_point(points1n[:, 0], points2n[:, 0],
P1, P2)
# Convert P2 from camera view to world view
P2_homogenous = np.linalg.inv(np.vstack([P2, [0, 0, 0, 1]]))
d2 = np.dot(P2_homogenous[:3, :4], d1)
if d1[2] > 0 and d2[2] > 0:
ind = i
P2 = np.linalg.inv(np.vstack([P2s[ind], [0, 0, 0, 1]]))[:3, :4]
tripoints3d = structure.linear_triangulation(points1n, points2n, P1, P2)
return tripoints3d, E, P1, P2
def read_directory(directory_name):
for filename in os.listdir(directory_name):
directory_name1 = "C:/yzn/picture/computer1"
def dino():
img1 = cv2.imread(directory_name + "/" + filename)
img2 = cv2.imread(directory_name1 + "/" + filename)
pts1, pts2 = features.find_correspondence_points(img1, img2)
points1 = processor.cart2hom(pts1)
points2 = processor.cart2hom(pts2)
fig, ax = plt.subplots(1, 2)
ax[0].autoscale_view('tight')
ax[0].imshow(cv2.cvtColor(img1, cv2.COLOR_BGR2RGB))
ax[0].plot(points1[0], points1[1], 'r.')
ax[1].autoscale_view('tight')
ax[1].imshow(cv2.cvtColor(img2, cv2.COLOR_BGR2RGB))
ax[1].plot(points2[0], points2[1], 'r.')
fig.show()
height, width, ch = img1.shape
intrinsic = np.array([ # for dino
[2360, 0, width / 2], [0, 2360, height / 2], [0, 0, 1]
])
return points1, points2, intrinsic
points1, points2, intrinsic = dino()
# Calculate essential matrix with 2d points.
# Result will be up to a scale
# First, normalize points
points1n = np.dot(np.linalg.inv(intrinsic), points1)
points2n = np.dot(np.linalg.inv(intrinsic), points2)
E = structure.compute_essential_normalized(points1n, points2n)
print('Computed essential matrix:', (-E / E[0][1]))
# Given we are at camera 1, calculate the parameters for camera 2
# Using the essential matrix returns 4 possible camera paramters
P1 = np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])
P2s = structure.compute_P_from_essential(E)
ind = -1
for i, P2 in enumerate(P2s):
# Find the correct camera parameters
d1 = structure.reconstruct_one_point(points1n[:, 0],
points2n[:, 0], P1, P2)
# Convert P2 from camera view to world view
P2_homogenous = np.linalg.inv(np.vstack([P2, [0, 0, 0, 1]]))
d2 = np.dot(P2_homogenous[:3, :4], d1)
if d1[2] > 0 and d2[2] > 0:
ind = i
P2 = np.linalg.inv(np.vstack([P2s[ind], [0, 0, 0, 1]]))[:3, :4]
#tripoints3d = structure.reconstruct_points(points1n, points2n, P1, P2)
tripoints3d = structure.linear_triangulation(points1n, points2n, P1,
P2)
print('3d ax', tripoints3d)
fig = plt.figure()
ax = fig.gca(projection='3d')
ax.plot(tripoints3d[0], tripoints3d[1], tripoints3d[2], 'b.')
ax.set_xlabel('x axis')
ax.set_ylabel('y axis')
ax.set_zlabel('z axis')
ax.view_init(elev=135, azim=90)
plt.show()
points1n = np.dot(np.linalg.inv(intrinsic), points1)
points2n = np.dot(np.linalg.inv(intrinsic), points2)
E = structure.compute_essential_normalized(points1n, points2n)
print('Computed essential matrix:', (-E / E[0][1]))
# True fundamental matrix F = K^-t E K^-1
true_F = np.dot(np.dot(np.linalg.inv(intrinsic).T, true_E),
np.linalg.inv(intrinsic))
F = structure.compute_fundamental_normalized(points1, points2)
print('True fundamental matrix:', true_F)
print('Computed fundamental matrix:', (F * true_F[2][2]))
# Given we are at camera 1, calculate the parameters for camera 2
# Using the essential matrix returns 4 possible camera paramters
P1 = np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])
P2s = structure.compute_P_from_essential(E)
ind = -1
for i, P2 in enumerate(P2s):
# Find the correct camera parameters
d1 = structure.reconstruct_one_point(points1n[:, 0], points2n[:, 0], P1,
P2)
P2_homogenous = extrinsic_from_camera_pose(P2)
d2 = np.dot(P2_homogenous[:3, :4], d1)
if d1[2] > 0 and d2[2] > 0:
ind = i
print('True pose of c2 wrt c1: ', H_c1_c2)
P2 = np.linalg.inv(np.vstack([P2s[ind], [0, 0, 0, 1]]))[:3, :4]
P2f = structure.compute_P_from_fundamental(F)