def plot_sfm_data_3d(sfm_data: SfmData,
ax: Axes,
max_plot_radius: float = 50) -> None:
"""Plot the camera poses and landmarks in 3D matplotlib plot.
Args:
sfm_data: SfmData object with camera and tracks.
ax: axis to plot on.
max_plot_radius: maximum distance threshold away from any camera for which a point
will be plotted
"""
# extract camera poses
camera_poses = []
for i in range(sfm_data.number_cameras()):
camera_poses.append(sfm_data.camera(i).pose())
plot_poses_3d(camera_poses, ax)
num_tracks = sfm_data.number_tracks()
# Restrict 3d points to some radius of camera poses
points_3d = np.array(
[list(sfm_data.track(j).point3()) for j in range(num_tracks)])
nearby_points_3d = comp_utils.get_points_within_radius_of_cameras(
camera_poses, points_3d, max_plot_radius)
# plot 3D points
for landmark in nearby_points_3d:
ax.plot(landmark[0], landmark[1], landmark[2], "g.", markersize=1)
def from_sfm_data(cls, sfm_data: gtsam.SfmData) -> "GtsfmData":
"""Initialize from gtsam.SfmData instance.
Args:
sfm_data: camera parameters and point tracks.
Returns:
A new GtsfmData instancee.
"""
num_images = sfm_data.numberCameras()
gtsfm_data = cls(num_images)
for i in range(num_images):
camera = sfm_data.camera(i)
gtsfm_data.add_camera(i, camera)
for j in range(sfm_data.numberTracks()):
gtsfm_data.add_track(sfm_data.track(j))
return gtsfm_data
def run(self, initial_data: SfmData) -> SfmResult:
"""Run the bundle adjustment by forming factor graph and optimizing using Levenberg–Marquardt optimization.
Args:
initial_data: initialized cameras, tracks w/ their 3d landmark from triangulation.
Results:
optimized camera poses, 3D point w/ tracks, and error metrics.
"""
logger.info(
f"Input: {initial_data.number_tracks()} tracks on {initial_data.number_cameras()} cameras\n"
)
# noise model for measurements -- one pixel in u and v
measurement_noise = gtsam.noiseModel.Isotropic.Sigma(
IMG_MEASUREMENT_DIM, 1.0)
# Create a factor graph
graph = gtsam.NonlinearFactorGraph()
# Add measurements to the factor graph
for j in range(initial_data.number_tracks()):
track = initial_data.track(j) # SfmTrack
# retrieve the SfmMeasurement objects
for m_idx in range(track.number_measurements()):
# i represents the camera index, and uv is the 2d measurement
i, uv = track.measurement(m_idx)
# note use of shorthand symbols C and P
graph.add(
GeneralSFMFactorCal3Bundler(uv, measurement_noise, C(i),
P(j)))
# Add a prior on pose x1. This indirectly specifies where the origin is.
graph.push_back(
gtsam.PriorFactorPinholeCameraCal3Bundler(
C(0),
initial_data.camera(0),
gtsam.noiseModel.Isotropic.Sigma(PINHOLE_CAM_CAL3BUNDLER_DOF,
0.1),
))
# Also add a prior on the position of the first landmark to fix the scale
graph.push_back(
gtsam.PriorFactorPoint3(
P(0),
initial_data.track(0).point3(),
gtsam.noiseModel.Isotropic.Sigma(POINT3_DOF, 0.1),
))
# Create initial estimate
initial = gtsam.Values()
i = 0
# add each PinholeCameraCal3Bundler
for cam_idx in range(initial_data.number_cameras()):
camera = initial_data.camera(cam_idx)
initial.insert(C(i), camera)
i += 1
j = 0
# add each SfmTrack
for t_idx in range(initial_data.number_tracks()):
track = initial_data.track(t_idx)
initial.insert(P(j), track.point3())
j += 1
# Optimize the graph and print results
try:
params = gtsam.LevenbergMarquardtParams()
params.setVerbosityLM("ERROR")
lm = gtsam.LevenbergMarquardtOptimizer(graph, initial, params)
result_values = lm.optimize()
except Exception as e:
logger.exception("LM Optimization failed")
return SfmResult(SfmData(), float("Nan"))
final_error = graph.error(result_values)
# Error drops from ~2764.22 to ~0.046
logger.info(f"initial error: {graph.error(initial):.2f}")
logger.info(f"final error: {final_error:.2f}")
# construct the results
optimized_data = values_to_sfm_data(result_values, initial_data)
sfm_result = SfmResult(optimized_data, final_error)
return sfm_result