Solves the Perspective-n-Point problem using the MLPnP algorithm with RANSAC. More...

Public Types
typedef Eigen::Vector3< g_type >	bearingVector_t
	A 3-vector of unit length used to describe landmark observations/bearings in camera frames (always expressed in camera frames).
typedef Buffer< bearingVector_t >	bearingVectors_t
	An array of bearing-vectors.
typedef Eigen::Matrix2< g_type >	cov2_mat_t
	A 2-matrix containing the 2D covariance information of a bearing Buffer.
typedef Eigen::Matrix3< g_type >	cov3_mat_t
	A 3-matrix containing the 3D covariance information of a bearing Buffer.
typedef Buffer< cov3_mat_t >	cov3_mats_t
	An array of 3D covariance matrices.
typedef Eigen::Vector4< g_type >	point4_t
	A homogeneous 3-vector describing a point in 3D-space.
typedef Eigen::Vector3< g_type >	point_t
	A 3-vector describing a point in 3D-space.
typedef Buffer< point4_t >	points4_t
	An array of homogeneous 3D-points.
typedef Buffer< point_t >	points_t
	An array of 3D-points.
typedef Eigen::Vector3< g_type >	rodrigues_t
	A 3-vector containing the rodrigues parameters of a rotation matrix.
typedef Eigen::Matrix3< g_type >	rotation_t
	A rotation matrix.
typedef Eigen::Matrix< g_type, 3, 4 >	transformation_t
	A 3x4 transformation matrix containing rotation $\mathbf{R}$ and translation $\mathbf{t}$ as follows: $\left( \begin{array}{cc} \mathbf{R} & \mathbf{t} \end{array} \right)$ .
typedef Eigen::Vector3< g_type >	translation_t
	A 3-vector describing a translation/camera position.

Public Member Functions
	MLPnPsolver (const Frame &F, const Buffer< MapPoint * > &vpMapPointMatches)
	Constructs the solver from a frame and map point matches.
bool	iterate (uint04 nIterations, bool &bNoMore, Buffer< bool > &vbInliers, uint04 &nInliers, Eigen::Matrix4f &Tout)
	Runs RANSAC iterations to estimate the camera pose.
void	SetRansacParameters (g_type probability=0.99, uint04 minInliers=8, uint04 maxIterations=300, uint04 minSet=6, float epsilon=0.4, float th2=5.991f)
	Sets RANSAC parameters.

Detailed Description

Solves the Perspective-n-Point problem using the MLPnP algorithm with RANSAC.

Given 2D-3D correspondences (bearing vectors and 3D points), estimates the camera pose using a maximum-likelihood formulation of PnP.

Definition at line 15 of file MLPnPsolver.h.

Constructor & Destructor Documentation

MLPnPsolver::MLPnPsolver	(	const Frame &	F,
		const Buffer< MapPoint * > &	vpMapPointMatches )

Constructs the solver from a frame and map point matches.

Parameters

[in]	F	The frame containing 2D observations.
[in]	vpMapPointMatches	Matched map points (nullptr for unmatched).

Runs RANSAC iterations to estimate the camera pose.

Parameters

[in]	nIterations	Number of iterations to run.
[out]	bNoMore	True if all iterations have been exhausted.
[out]	vbInliers	Inlier flags per correspondence.
[out]	nInliers	Number of inliers found.
[out]	Tout	The estimated 4x4 transformation matrix.

void MLPnPsolver::SetRansacParameters	(	g_type	probability = 0.99,
		uint04	minInliers = 8,
		uint04	maxIterations = 300,
		uint04	minSet = 6,
		float	epsilon = 0.4,
		float	th2 = 5.991f )

Sets RANSAC parameters.

Parameters

[in]	probability	Success probability.
[in]	minInliers	Minimum inlier count.
[in]	maxIterations	Maximum RANSAC iterations.
[in]	minSet	Minimum sample set size.
[in]	epsilon	Expected inlier ratio.
[in]	th2	Chi-squared threshold for inlier classification.

The documentation for this class was generated from the following file: