Back to list of all libraries | See all modules
mrpt-scanmatching
Under this name we find functions in charge of solving the optimization problem of aligning a set of correspondences, both in 2D and in 3D. Note that this does not includes the iterative ICP algorithm (see mrpt::slam::CICP), included in another library (mrpt-slam).
Namespaces | |
mrpt::scanmatching | |
A set of scan matching-related static functions. | |
Functions | |
double SCANMATCHING_IMPEXP | mrpt::scanmatching::HornMethod (const vector_double &inPoints, vector_double &outQuat, bool forceScaleToUnity=false) |
This function implements the Horn method for computing the change in pose between two coordinate systems. More... | |
double SCANMATCHING_IMPEXP | mrpt::scanmatching::HornMethod (const vector_double &inPoints, mrpt::poses::CPose3DQuat &outQuat, bool forceScaleToUnity=false) |
bool SCANMATCHING_IMPEXP | mrpt::scanmatching::leastSquareErrorRigidTransformation6D (const TMatchingPairList &in_correspondences, CPose3DQuat &out_transformation, double &out_scale, const bool forceScaleToUnity=false) |
This method provides the closed-form solution of absolute orientation using unit quaternions to a set of over-constrained correspondences for finding the 6D rigid transformation between two cloud of 3D points. More... | |
bool SCANMATCHING_IMPEXP | mrpt::scanmatching::leastSquareErrorRigidTransformation6D (const TMatchingPairList &in_correspondences, CPose3D &out_transformation, double &out_scale, const bool forceScaleToUnity=false) |
This method provides the closed-form solution of absolute orientation using unit quaternions to a set of over-constrained correspondences for finding the 6D rigid transformation between two cloud of 3D points. More... | |
bool SCANMATCHING_IMPEXP | mrpt::scanmatching::leastSquareErrorRigidTransformation6DRANSAC (const TMatchingPairList &in_correspondences, CPose3D &out_transformation, double &out_scale, vector_int &out_inliers_idx, const unsigned int ransac_minSetSize=5, const unsigned int ransac_nmaxSimulations=50, const double ransac_maxSetSizePct=0.7, const bool forceScaleToUnity=false) |
This method provides the closed-form solution of absolute orientation using unit quaternions to a set of over-constrained correspondences for finding the 6D rigid transformation between two cloud of 3D points using RANSAC. More... | |
bool SCANMATCHING_IMPEXP | mrpt::scanmatching::leastSquareErrorRigidTransformation (TMatchingPairList &in_correspondences, CPose2D &out_transformation, CMatrixDouble33 *out_estimateCovariance=NULL) |
This method provides the basic least-square-error solution to a set of over-constrained correspondences for finding the (x,y,phi) rigid transformation between two planes. More... | |
bool SCANMATCHING_IMPEXP | mrpt::scanmatching::leastSquareErrorRigidTransformation (TMatchingPairList &in_correspondences, CPosePDFGaussian &out_transformation) |
This method provides the basic least-square-error solution to a set of over-constrained correspondences for finding the (x,y,phi) rigid transformation between two planes. More... | |
void SCANMATCHING_IMPEXP | mrpt::scanmatching::robustRigidTransformation (TMatchingPairList &in_correspondences, poses::CPosePDFSOG &out_transformation, float normalizationStd, unsigned int ransac_minSetSize=3, unsigned int ransac_maxSetSize=20, float ransac_mahalanobisDistanceThreshold=3.0f, unsigned int ransac_nSimulations=0, TMatchingPairList *out_largestSubSet=NULL, bool ransac_fuseByCorrsMatch=true, float ransac_fuseMaxDiffXY=0.01f, float ransac_fuseMaxDiffPhi=DEG2RAD(0.1f), bool ransac_algorithmForLandmarks=true, double probability_find_good_model=0.999, unsigned int ransac_min_nSimulations=1500, const bool verbose=false, double max_rmse_to_end=0) |
This method implements a RANSAC-based robust estimation of the rigid transformation between two planar frames of references, returning a probability distribution over all the posibilities as a Sum of Gaussians. More... | |
double SCANMATCHING_IMPEXP mrpt::scanmatching::HornMethod | ( | const vector_double & | inPoints, |
vector_double & | outQuat, | ||
bool | forceScaleToUnity = false |
||
) |
This function implements the Horn method for computing the change in pose between two coordinate systems.
[in] | inPoints | A vector containing the coordinates of the input points in the format: [x11 y11 z11, x12 y12 z12, x21 y21 z21, x22 y22 z22, x31 y31 z31, x32 y32 z32, ... ] where [xi1 yi1 zi1] and [xi2 yi2 zi2] represent the i-th pair of corresponding 3D points in the two coordinate systems "1" and "2" |
[out] | outQuat | A 7D vector containing the traslation and rotation (in a quaternion form) which indicates the change in pose of system "2" wrt "1". |
[in] | forceScaleToUnity | Whether or not force the scale employed to rotate the coordinate systems to one (rigid transformation) |
double SCANMATCHING_IMPEXP mrpt::scanmatching::HornMethod | ( | const vector_double & | inPoints, |
mrpt::poses::CPose3DQuat & | outQuat, | ||
bool | forceScaleToUnity = false |
||
) |
This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.
bool SCANMATCHING_IMPEXP mrpt::scanmatching::leastSquareErrorRigidTransformation | ( | TMatchingPairList & | in_correspondences, |
CPose2D & | out_transformation, | ||
CMatrixDouble33 * | out_estimateCovariance = NULL |
||
) |
This method provides the basic least-square-error solution to a set of over-constrained correspondences for finding the (x,y,phi) rigid transformation between two planes.
The optimal transformation q fulfills:
in_correspondences | The set of correspondences. |
out_transformation | The pose that minimizes the mean-square-error between all the correspondences. |
out_estimateCovariance | If provided (!=NULL) this will contain on return a 3x3 covariance matrix with the NORMALIZED optimal estimate uncertainty. This matrix must be multiplied by , the variance of matched points in and (see paper http://www.mrpt.org/Paper:Occupancy_Grid_Matching) |
Raises | a std::exception if the list "in_correspondences" has not a minimum of two correspondences. |
bool SCANMATCHING_IMPEXP mrpt::scanmatching::leastSquareErrorRigidTransformation | ( | TMatchingPairList & | in_correspondences, |
CPosePDFGaussian & | out_transformation | ||
) |
This method provides the basic least-square-error solution to a set of over-constrained correspondences for finding the (x,y,phi) rigid transformation between two planes.
The optimal transformation q fulfills:
in_correspondences | The set of correspondences. |
out_transformation | The pose that minimizes the mean-square-error between all the correspondences. |
out_estimateCovariance | If provided (!=NULL) this will contain on return a 3x3 covariance matrix with the NORMALIZED optimal estimate uncertainty. This matrix must be multiplied by , the variance of matched points in and (see paper http://www.mrpt.org/Paper:Occupancy_Grid_Matching) |
Raises | a std::exception if the list "in_correspondences" has not a minimum of two correspondences. |
bool SCANMATCHING_IMPEXP mrpt::scanmatching::leastSquareErrorRigidTransformation6D | ( | const TMatchingPairList & | in_correspondences, |
CPose3DQuat & | out_transformation, | ||
double & | out_scale, | ||
const bool | forceScaleToUnity = false |
||
) |
This method provides the closed-form solution of absolute orientation using unit quaternions to a set of over-constrained correspondences for finding the 6D rigid transformation between two cloud of 3D points.
The output 3D pose is computed using the method described in "Closed-form solution of absolute orientation using unit quaternions", BKP Horn, Journal of the Optical Society of America, 1987.
in_correspondences | The set of correspondences in TMatchingPairList form ("this" and "other"). |
out_transformation | The change in pose (CPose3DQuat) of the "other" reference system wrt "this" reference system which minimizes the mean-square-error between all the correspondences. |
Raises | a std::exception if the list "in_correspondences" has not a minimum of three correspondences. |
bool SCANMATCHING_IMPEXP mrpt::scanmatching::leastSquareErrorRigidTransformation6D | ( | const TMatchingPairList & | in_correspondences, |
CPose3D & | out_transformation, | ||
double & | out_scale, | ||
const bool | forceScaleToUnity = false |
||
) |
This method provides the closed-form solution of absolute orientation using unit quaternions to a set of over-constrained correspondences for finding the 6D rigid transformation between two cloud of 3D points.
The output 3D pose is computed using the method described in "Closed-form solution of absolute orientation using unit quaternions", BKP Horn, Journal of the Optical Society of America, 1987.
in_correspondences | The set of correspondences. |
out_transformation | The change in pose (CPose3DQuat) of the "other" reference system wrt "this" reference system which minimizes the mean-square-error between all the correspondences. |
Raises | a std::exception if the list "in_correspondences" has not a minimum of two correspondences. |
bool SCANMATCHING_IMPEXP mrpt::scanmatching::leastSquareErrorRigidTransformation6DRANSAC | ( | const TMatchingPairList & | in_correspondences, |
CPose3D & | out_transformation, | ||
double & | out_scale, | ||
vector_int & | out_inliers_idx, | ||
const unsigned int | ransac_minSetSize = 5 , |
||
const unsigned int | ransac_nmaxSimulations = 50 , |
||
const double | ransac_maxSetSizePct = 0.7 , |
||
const bool | forceScaleToUnity = false |
||
) |
This method provides the closed-form solution of absolute orientation using unit quaternions to a set of over-constrained correspondences for finding the 6D rigid transformation between two cloud of 3D points using RANSAC.
The output 3D pose is computed using the method described in "Closed-form solution of absolute orientation using unit quaternions", BKP Horn, Journal of the Optical Society of America, 1987. If supplied, the output covariance matrix is computed using... TODO
in_correspondences | The set of correspondences. |
out_transformation | The pose that minimizes the mean-square-error between all the correspondences. |
out_scale | The estimated scale of the rigid transformation (should be very close to 1.0) |
out_inliers_idx | Indexes within the "in_correspondences" list which corresponds with inliers |
ransac_minSetSize | The minimum amount of points in the set |
ransac_nmaxSimulations | The maximum number of iterations of the RANSAC algorithm |
ransac_maxSetSizePct | The (minimum) assumed percent (0.0 - 1.0) of the input set to be considered as inliers |
Raises | a std::exception if the list "in_correspondences" has not a minimum of two correspondences. |
void SCANMATCHING_IMPEXP mrpt::scanmatching::robustRigidTransformation | ( | TMatchingPairList & | in_correspondences, |
poses::CPosePDFSOG & | out_transformation, | ||
float | normalizationStd, | ||
unsigned int | ransac_minSetSize = 3 , |
||
unsigned int | ransac_maxSetSize = 20 , |
||
float | ransac_mahalanobisDistanceThreshold = 3.0f , |
||
unsigned int | ransac_nSimulations = 0 , |
||
TMatchingPairList * | out_largestSubSet = NULL , |
||
bool | ransac_fuseByCorrsMatch = true , |
||
float | ransac_fuseMaxDiffXY = 0.01f , |
||
float | ransac_fuseMaxDiffPhi = DEG2RAD(0.1f) , |
||
bool | ransac_algorithmForLandmarks = true , |
||
double | probability_find_good_model = 0.999 , |
||
unsigned int | ransac_min_nSimulations = 1500 , |
||
const bool | verbose = false , |
||
double | max_rmse_to_end = 0 |
||
) |
This method implements a RANSAC-based robust estimation of the rigid transformation between two planar frames of references, returning a probability distribution over all the posibilities as a Sum of Gaussians.
The technique was described in the paper:
This works are follows:
For more details refer to the tutorial on scan matching methods. NOTE:
If ransac_fuseByCorrsMatch=true (the default), the weight of Gaussian modes will be increased when an exact match in the subset of correspondences for the modes is found. Otherwise, an approximate method is used as test by just looking at the resulting X,Y,PHI means (Threshold in this case are: ransac_fuseMaxDiffXY, ransac_fuseMaxDiffPhi).
[in] | max_rmse_to_end | Stop searching for solutions when the RMSE of one solution is below this threshold. Special value "0" means "auto", which employs "2*normalizationStd". |
Raises | a std::exception if the list "in_correspondences" has not a minimum of two correspondences. |
Page generated by Doxygen 1.8.14 for MRPT 1.1.0 SVN: at lun oct 28 00:54:49 CET 2019 | Hosted on: |