opentl::modelprojection::PoseEstimation Class Reference

Pose estimation from point correspondences. This class implements pose estimation from point and/or line correspondences. It uses a two-step procedure: 1 - Initialize pose parameters, using initPose methods from the Pose abstraction 2 - Runs (if required) a Gauss-Newton nonlinear optimization, in order to refine the estimate. More...

List of all members.

Public Member Functions
	PoseEstimation (modelprojection::Warp *warp)
	Constructor.
int	poseFromPoints (const opentl::core::State &state, const math::Matrix &bodyPoints, const math::Matrix &screenPoints, int cameraIdx=0, const math::Vector *linkIdx=NULL, bool useGaussNewton=false, double errCov=1.0)
	Compute the pose from a set of point correspondences (model-image), eventually using a Gauss-Newton refinement.
	~PoseEstimation ()
	Destructor.

---

Detailed Description

Pose estimation from point correspondences. This class implements pose estimation from point and/or line correspondences. It uses a two-step procedure: 1 - Initialize pose parameters, using initPose methods from the Pose abstraction 2 - Runs (if required) a Gauss-Newton nonlinear optimization, in order to refine the estimate.

Author:

panin@in.tum.de

---

Constructor & Destructor Documentation

opentl::modelprojection::PoseEstimation::PoseEstimation

(

modelprojection::Warp *

warp

)

Constructor.

opentl::modelprojection::PoseEstimation::~PoseEstimation

(

)

Destructor.

---

Member Function Documentation

int opentl::modelprojection::PoseEstimation::poseFromPoints	(	const opentl::core::State &	state,
		const math::Matrix &	bodyPoints,
		const math::Matrix &	screenPoints,
		int	cameraIdx = 0,
		const math::Vector *	linkIdx = NULL,
		bool	useGaussNewton = false,
		double	errCov = 1.0
	)

Compute the pose from a set of point correspondences (model-image), eventually using a Gauss-Newton refinement.

Parameters:

	state	Input/output pose to be estimated (within a State vector)
	bodyPoints	Input points in object space (3xN) or (2xN), for each target
	screenPoints	Input corresponding points in image space (2xN), for each target
	cameraIdx	Camera ID for all targets (default = 0)
	linkIdx	Input Link ID of each point, for each target (for articulated Poses only) - default = NULL
	useGaussNewton	Flag, whether to refine the estimate with Gauss-Newton - default = false
	errCov	Covariance of point re-projection errors - deafult = 1.0

Returns:

Error condition (0 = OK)