mrpt::vision::pnp::CPnP Class Reference

Detailed Description

This class is used for Pose estimation from a known landmark using a monocular camera.

The toolkit comprises of state of the art PnP (Perspective n Point) algorithms

The Python Bindings pnp_perf_comp.py can be used to generate performance comparison using standard tests between the different algorithms.

Performance comparison Results

Author

Chandra Mangipudi

Date

17/08/16

Definition at line 47 of file pnp_algos.h.

#include <mrpt/vision/pnp_algos.h>

Public Member Functions
bool	dls (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[3] Direct Least Squares (DLS) - PnP : Algorithm formulates position as a function of rotation. More...
bool	epnp (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[8] Efficient-PnP (EPnP) - Algorithm takes 4 control points based on n points and uses the control points to compute the pose More...
bool	upnp (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[5] Unified-PnP (UPnP) : Algorithm to compute pose from unknown camera intrinsic matrix More...
bool	p3p (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[4] P3P - A closed form solution for n=3, 2D-3D correspondences More...
bool	rpnp (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[6] Robust (R-PnP)- A closed form solution with intermediate P3P computations More...
bool	ppnp (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[2] Procrustes-PnP(PPnP) Algorithm : Iterative SVD based algorithm More...
bool	posit (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[1] Pose from Orthogoanlity and Scaling :Iterative (POSIT) - A Geometric algorithm to compute scale and orthogonality independently More...
bool	lhm (const Eigen::Ref< Eigen::MatrixXd > obj_pts, const Eigen::Ref< Eigen::MatrixXd > img_pts, int n, const Eigen::Ref< Eigen::MatrixXd > cam_intrinsic, Eigen::Ref< Eigen::MatrixXd > pose_mat)
	[7] Lu-Hager-Mjolsness(LHM)-PnP algorithm : Iterative algorithm to reduce object space error More...

Member Function Documentation

dls()

bool mrpt::vision::pnp::CPnP::dls	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[3] Direct Least Squares (DLS) - PnP : Algorithm formulates position as a function of rotation.

Use Cayley's rotation theorem to represent the rotation using parameters ($s_1, s_2, s_3$). Solve the rotation using multi-variate polynomial expressions

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 (only 4 points used) array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 (only 4 points used) array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 45 of file pnp_algos.cpp.

References mrpt::vision::pnp::dls::compute_pose().

epnp()

bool mrpt::vision::pnp::CPnP::epnp	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[8] Efficient-PnP (EPnP) - Algorithm takes 4 control points based on n points and uses the control points to compute the pose

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 (only 4 points used) array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 (only 4 points used) array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 110 of file pnp_algos.cpp.

References mrpt::vision::pnp::epnp::compute_pose().

lhm()

bool mrpt::vision::pnp::CPnP::lhm	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[7] Lu-Hager-Mjolsness(LHM)-PnP algorithm : Iterative algorithm to reduce object space error

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 array [p_x p_y p_z]
[in]	img_pts	Image points in pixels nX3 array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 442 of file pnp_algos.cpp.

References mrpt::vision::pnp::lhm::compute_pose(), and R.

p3p()

bool mrpt::vision::pnp::CPnP::p3p	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[4] P3P - A closed form solution for n=3, 2D-3D correspondences

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 (only 4 points used) array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 (only 4 points used) array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 239 of file pnp_algos.cpp.

References R.

posit()

bool mrpt::vision::pnp::CPnP::posit	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[1] Pose from Orthogoanlity and Scaling :Iterative (POSIT) - A Geometric algorithm to compute scale and orthogonality independently

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 390 of file pnp_algos.cpp.

References R.

ppnp()

bool mrpt::vision::pnp::CPnP::ppnp	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[2] Procrustes-PnP(PPnP) Algorithm : Iterative SVD based algorithm

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 339 of file pnp_algos.cpp.

References R.

rpnp()

bool mrpt::vision::pnp::CPnP::rpnp	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[6] Robust (R-PnP)- A closed form solution with intermediate P3P computations

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 (only 4 points used) array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 (only 4 points used) array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 290 of file pnp_algos.cpp.

References R.

upnp()

bool mrpt::vision::pnp::CPnP::upnp	(	const Eigen::Ref< Eigen::MatrixXd >	obj_pts,
		const Eigen::Ref< Eigen::MatrixXd >	img_pts,
		int	n,
		const Eigen::Ref< Eigen::MatrixXd >	cam_intrinsic,
		Eigen::Ref< Eigen::MatrixXd >	pose_mat
	)

[5] Unified-PnP (UPnP) : Algorithm to compute pose from unknown camera intrinsic matrix

Parameters

[in]	obj_pts	Object points in Camera Co-ordinate system {C} nX3 (only 4 points used) array [p_x, p_y, p_z]
[in]	img_pts	Image points in pixels nX3 (only 4 points used) array containing pixel data from camera [u, v, 1]
[in]	n	number of 2D-3D correspondences
[in]	cam_intrinsic	Camera Intrinsic matrix
[out]	pose_mat	Output pose vector 6X1, pose_mat[0:2]-> Translation, pose_mat[3:5] -> Quaternion vector component

Returns

success flag

Definition at line 174 of file pnp_algos.cpp.

References mrpt::vision::pnp::upnp::compute_pose().