Sensors¶

General¶

All sensors derive from a common base class. This base class is not available from Python.

class kontiki.sensors.Sensor¶

relative_pose¶: A tuple (q, p) where p is the relative position, and q the relative orientation between the sensor coordinate frame, and the trajectory coordinate frame such that X_sensor = R * X_world + p. q is a unit quaternion representing the orientation R. Defaults to the identity orientation, and zero, respectively.

time_offset¶: The time offset d between a reference time frame, and the sensor. Any function foo(t) of the sensor is instead called as foo(t + d). Defaults to 0.

max_time_offset¶: When optimizing the time offset, this is the maximum deviation from 0.

from_trajectory(X : ndarray) → ndarray¶: Moves the 3D point X from the trajectory to the sensor coordinate frame, by applying the relative pose.

to_trajectory(X : ndarray) → ndarray¶: Moves the 3D point X from the sensor to the trajectory coordinate frame, by applying the relative pose.

relative_orientation_locked¶: If True, the orientation part of the relative pose can not be changed during optimization. Defaults to True.

relative_position_locked¶: If True, the translation part of the relative pose can not be changed during optimization. Defaults to True.

time_offset_locked¶: If True, the time offset can not be changed during optimization. Defaults to True.

Cameras¶

All cameras share the following (unavailable) base class

class kontiki.sensors.Camera¶

rows¶: The number of image rows

cols¶: The number of image columns

readout¶: Rolling shutter readout time. Set to 0 for global shutter.

project(X : ndarray) → ndarray¶: Projects the 3D point X in camera coordinates, to its 2D image plane location using the camera projection model.

Note

This does not apply the relative pose of the camera

unproject(y : ndarray) → ndarray¶: Apply inverse projection on image (2D) location y, to get a 3D point X=(x, y, z), where z=1.

Note

This does not apply the relative pose of the camera

to_trajectory(X : ndarray) → ndarray¶: Moves the 3D point X in camera coordinates t

Camera classes¶

class kontiki.sensors.PinholeCamera¶

Standard pinhole camera

camera_matrix¶: The 3x3 internal camera calibration matrix

class kontiki.sensors.AtanCamera¶

Arc-tan camera model for wide-angle cameras

Implementation of the 3-parameter FOV model of [Devernay2001].

Subclass of PinholeCamera.

References

[Devernay2001]

Devernay, F. and Faugeras, O. Straight lines have to be straight: Automatic calibration and removal of distortion from scenes of structured environments Machie Vision and Applications, Vol 13., 2001

gamma¶: Distortion parameter

wc¶: Distortion center

IMUs¶

All inertial measurement units share the following (unavailable) base class:

class kontiki.sensors.Imu¶

accelerometer(trajectory : Trajectory, t : float) → ndarray¶: A single accelerometer measurement at time t, using the specified trajectory.

gyroscope(trajectory : Trajectory, t : float) → ndarray¶: A single gyroscope measurement at time t, using the specified trajectory.

IMU classes¶

class kontiki.sensors.BasicImu¶

Basic IMU

This IMU model produces ideal IMU measurements.

class kontiki.sensors.ConstantBiasImu¶

IMU with constant biases

Both biases are modeled as constants.

accelerometer_bias¶: Accelerometer bias

accelerometer_bias_locked¶: If True (default), then the accelerometer bias is not changed during optimization.

gyroscope_bias¶: Gyroscope bias

gyroscope_bias_locked¶: If True (default), then the gyrocope bias is not changed during optimization.