Ceres¶

Bundle Adjust Algorithm¶

class kwiver::arrows::ceres::bundle_adjust¶

A class for bundle adjustment of feature tracks using Ceres.

Inherits from kwiver::vital::algorithm_impl< bundle_adjust, vital::algo::bundle_adjust >

Public Functions

bundle_adjust()¶: Constructor.

~bundle_adjust()¶: Destructor.

config_block_sptr get_configuration() const¶: Get this algorithm’s configuration block .

void set_configuration(vital::config_block_sptr config)¶: Set this algorithm’s properties via a config block.

bool check_configuration(vital::config_block_sptr config) const¶: Check that the algorithm’s currently configuration is valid.

void optimize(vital::camera_map_sptr &cameras, vital::landmark_map_sptr &landmarks, vital::feature_track_set_sptr tracks, vital::video_metadata_map_sptr metadata = nullptr) const¶

Optimize the camera and landmark parameters given a set of feature tracks.

Optimize the camera and landmark parameters given a set of tracks.

Parameters

cameras: the cameras to optimize
landmarks: the landmarks to optimize
tracks: the feature tracks to use as constraints
metadata: the frame metadata to use as constraints

void set_callback(callback_t cb)¶: Set a callback function to report intermediate progress.

bool trigger_callback()¶: This function is called by a Ceres callback to trigger a kwiver callback.

class priv¶

Private implementation class.

Inherits from kwiver::arrows::ceres::solver_options, kwiver::arrows::ceres::camera_options

Public Functions

priv()¶: Constructor.

Public Members

bool verbose¶: verbose output

LossFunctionType loss_function_type¶: the robust loss function type to use

double loss_function_scale¶: the scale of the loss function

camera_map::map_camera_t cams¶: the input cameras to update in place

landmark_map::map_landmark_t lms¶: the input landmarks to update in place

std::map<track_id_t, std::vector<double>> landmark_params¶: a map from track id to landmark parameters

std::map<frame_id_t, std::vector<double>> camera_params¶: a map from frame number to extrinsic parameters

std::vector<std::vector<double>> camera_intr_params¶: vector of unique camera intrinsic parameters

std::map<frame_id_t, unsigned int> frame_to_intr_map¶: a map from frame number to index of unique camera intrinsics in camera_intr_params

StateCallback ceres_callback¶: the ceres callback class

vital::logger_handle_t m_logger¶: Logger handle.

Optimize Cameras Algorithm¶

class kwiver::arrows::ceres::optimize_cameras¶

A class for optimization of camera paramters using Ceres.

Inherits from kwiver::vital::algorithm_impl< optimize_cameras, vital::algo::optimize_cameras >

Public Functions

optimize_cameras()¶: Constructor.

~optimize_cameras()¶: Destructor.

optimize_cameras(const optimize_cameras &other)¶: Copy Constructor.

config_block_sptr get_configuration() const¶: Get this algorithm’s configuration block .

void set_configuration(vital::config_block_sptr config)¶: Set this algorithm’s properties via a config block.

bool check_configuration(vital::config_block_sptr config) const¶: Check that the algorithm’s currently configuration is valid.

void optimize(kwiver::vital::camera_map_sptr &cameras, kwiver::vital::feature_track_set_sptr tracks, kwiver::vital::landmark_map_sptr landmarks, kwiver::vital::video_metadata_map_sptr metadata = nullptr) const¶

Optimize camera parameters given sets of landmarks and feature tracks.

We only optimize cameras that have associating tracks and landmarks in the given maps. The default implementation collects the corresponding features and landmarks for each camera and calls the single camera optimize function.

Exceptions

invalid_value: When one or more of the given pointer is Null.

Parameters

cameras: Cameras to optimize.
tracks: The feature tracks to use as constraints.
landmarks: The landmarks the cameras are viewing.
metadata: The optional metadata to constrain the optimization.

void optimize(vital::camera_sptr &camera, const std::vector<vital::feature_sptr> &features, const std::vector<vital::landmark_sptr> &landmarks, kwiver::vital::video_metadata_vector metadata = kwiver::vital::video_metadata_vector()) const¶

Optimize a single camera given corresponding features and landmarks.

This function assumes that 2D features viewed by this camera have already been put into correspondence with 3D landmarks by aligning them into two parallel vectors

Parameters

camera: The camera to optimize.
features: The vector of features observed by camera to use as constraints.
landmarks: The vector of landmarks corresponding to features.
metadata: The optional metadata to constrain the optimization.

class priv¶

Private implementation class.

Inherits from kwiver::arrows::ceres::solver_options, kwiver::arrows::ceres::camera_options

Public Functions

priv()¶: Constructor.

Public Members

bool verbose¶: verbose output

LossFunctionType loss_function_type¶: the robust loss function type to use

double loss_function_scale¶: the scale of the loss function

vital::logger_handle_t m_logger¶: Logger handle.

Camera Position Smoothness Class¶

class kwiver::arrows::ceres::camera_position_smoothness¶

Ceres camera smoothness functor.

Public Functions

camera_position_smoothness(const double smoothness)¶: Constructor.

template <typename T>

bool operator()(const T *const prev_pose, const T *const curr_pose, const T *const next_pose, T *residuals) const¶

Position smoothness error functor for use in Ceres.

Parameters

prev_pos: Camera pose data block at previous time
curr_pos: Camera pose data block at current time
next_pos: Camera pose data block at next time
residuals: Camera pose blocks contain 6 parameters: 3 for rotation(angle axis), 3 for center Only the camera centers are used in this function to penalize the difference between current position and the average between previous and next positions.

Public Static Functions

ceres::CostFunction *create(const double s)¶: Cost function factory.

Camera Limit Forward Motion Class¶

class kwiver::arrows::ceres::camera_limit_forward_motion¶

Ceres camera limit forward motion functor.

This class is to reglarize camera motion to minimize the amount of motion in the camera looking direction. This is useful with zoom lenses at long focal lengths where distance and zoom are ambiguous. Adding this constraint allows the optimization to prefer fast zoom changes over fast position change.

Public Functions

camera_limit_forward_motion(const double scale)¶: Constructor.

template <typename T>

bool operator()(const T *const pose1, const T *const pose2, T *residuals) const¶

Camera forward motion error functor for use in Ceres.

Parameters

pose1: Camera pose data block at time 1
pose2: Camera pose data block at time 2
residuals: Camera pose blocks contain 6 parameters: 3 for rotation(angle axis), 3 for center

Public Members

double scale_¶: the magnitude of this constraint

Public Static Functions

ceres::CostFunction *create(const double s)¶: Cost function factory.

Distortion Poly Radial Class¶

class kwiver::arrows::ceres::distortion_poly_radial¶

Class to hold to distortion function and traits.

Public Static Functions

template <typename T>

static void apply(const T *dist_coeffs, const T *source_xy, T *distorted_xy)¶

Function to apply polynomial radial distortion.

Parameters

dist_coeffs: radial distortion coefficients (2)
source_xy: 2D point in normalized image coordinates
distorted_xy: 2D point in distorted normalized image coordinates

Distortion Poly Radial Tangential Class¶

class kwiver::arrows::ceres::distortion_poly_radial_tangential¶

Class to hold to distortion function and traits.

Public Static Functions

template <typename T>

static void apply(const T *dist_coeffs, const T *source_xy, T *distorted_xy)¶

Function to apply polynomial radial and tangential distortion.

Parameters

dist_coeffs: radial (3) and tangential (2) distortion coefficients
source_xy: 2D point in normalized image coordinates
distorted_xy: 2D point in distorted normalized image coordinates

Distortion Ratpoly Radial Tangential Class¶

class kwiver::arrows::ceres::distortion_ratpoly_radial_tangential¶

Class to hold to distortion function and traits.

Public Static Functions

template <typename T>

static void apply(const T *dist_coeffs, const T *source_xy, T *distorted_xy)¶

Function to apply rational polynomial radial and tangential distortion.

Parameters

dist_coeffs: radial (6) and tangential (2) distortion coefficients
source_xy: 2D point in normalized image coordinates
distorted_xy: 2D point in distorted normalized image coordinates

Create Cost Func Factory¶

KWIVER_ALGO_CERES_EXPORT::ceres::CostFunction *kwiver::arrows::ceres::create_cost_func(LensDistortionType ldt, double x, double y)¶: Factory to create Ceres cost functions for each lens distortion type.