Class ICP

Nested Relationships

Nested Types

Inheritance Relationships

Base Types

  • public mrpt::system::COutputLogger

  • public mrpt::rtti::CObject

Derived Type

Class Documentation

class ICP : public mrpt::system::COutputLogger, public mrpt::rtti::CObject

Generic ICP algorithm container: builds a custom ICP pipeline by selecting algorithm and parameter for each stage.

The main API entry point is align().

A convenient way to create an ICP pipeline instance is using a YAML configuration file and calling mp2p_icp::icp_pipeline_from_yaml().

Todo:

Add pipeline picture.

Several solvers may exists, but the output from the first one returning true will be used. This is by design, to enable different solver algorithms depending on the ICP iteration or the type of geometric entities.

Subclassed by mp2p_icp::ICP_LibPointmatcher

Module: Solver instances

using solver_list_t = std::vector<mp2p_icp::Solver::Ptr>
void initialize_solvers(const mrpt::containers::yaml &params)

Create and configure one or more “Solver” modules from YAML-like config block. Config must be a sequence of one or more entries, each with a class and a params dictionary entries.

Read the comments for ICP on the possible existence of more than one solver.

Example: 

- class: mp2p_icp::Solver_Horn
  params:
   # Parameters depend on the particular class
   # none

Alternatively, the objects can be directly created via solvers().

See also

mp2p_icp::icp_pipeline_from_yaml()

inline const solver_list_t &solvers() const
inline solver_list_t &solvers()
static void initialize_solvers(const mrpt::containers::yaml &params, ICP::solver_list_t &lst)
static bool run_solvers(const solver_list_t &solvers, const Pairings &pairings, OptimalTF_Result &out, const SolverContext &sc = {})

Runs a set of solvers.

Module: QualityEvaluator instances

using quality_eval_list_t = std::vector<QualityEvaluatorEntry>
void initialize_quality_evaluators(const mrpt::containers::yaml &params)

Create and configure one or more “QualityEvaluator” modules from YAML-like config block. Config must be a sequence of one or more entries, each with a class and a params dictionary entries.

Example: 

- class: mp2p_icp::QualityEvaluator_PairedRatio
  weight: 1.0  # (Optional if only one quality evaluator is defined)
  params:
   # Parameters depend on the particular class
   # xxx: yyyy

Alternatively, the objects can be directly created via matchers().

See also

mp2p_icp::icp_pipeline_from_yaml()

inline const quality_eval_list_t &quality_evaluators() const
inline quality_eval_list_t &quality_evaluators()
static void initialize_quality_evaluators(const mrpt::containers::yaml &params, quality_eval_list_t &lst)
static double evaluate_quality(const quality_eval_list_t &evaluators, const metric_map_t &pcGlobal, const metric_map_t &pcLocal, const mrpt::poses::CPose3D &localPose, const Pairings &finalPairings)

Module: Matcher instances

void initialize_matchers(const mrpt::containers::yaml &params)

Create and configure one or more “Match” modules from YAML-like config block. Config must be a sequence of one or more entries, each with a class and a params dictionary entries.

Example: 

- class: mp2p_icp::Matcher_Points_DistanceThreshold
  params:
   # Parameters depend on the particular class
   threshold: 1.0

Alternatively, the objects can be directly created via matchers().

See also

mp2p_icp::icp_pipeline_from_yaml()

inline const matcher_list_t &matchers() const
inline matcher_list_t &matchers()
static void initialize_matchers(const mrpt::containers::yaml &params, matcher_list_t &lst)

Public Types

using iteration_hook_t = std::function<IterationHook_Output(const IterationHook_Input&)>

Public Functions

virtual void align(const metric_map_t &pcLocal, const metric_map_t &pcGlobal, const mrpt::math::TPose3D &initialGuessLocalWrtGlobal, const Parameters &p, Results &result, const std::optional<mrpt::poses::CPose3DPDFGaussianInf> &prior = std::nullopt, const mrpt::optional_ref<LogRecord> &outputDebugInfo = std::nullopt)

Register (align) two point clouds (possibly after having been preprocessed to extract features, etc.) and returns the relative pose of pcLocal with respect to pcGlobal.

inline void attachToParameterSource(ParameterSource &source)
inline virtual void initialize_derived(const mrpt::containers::yaml &p)

For whole-ICP overriden classes (e.g. external ICP library wrappers), initialize those external libraries with these parameters. Invoked from mp2p_icp::icp_pipeline_from_yaml().

See also

mp2p_icp::icp_pipeline_from_yaml()

inline void setIterationHook(const iteration_hook_t &ih)
inline const mrpt::system::CTimeLogger &profiler() const
inline mrpt::system::CTimeLogger &profiler()

Protected Attributes

solver_list_t solvers_
matcher_list_t matchers_
quality_eval_list_t quality_evaluators_ = {{QualityEvaluator_PairedRatio::Create(), 1.0}}
iteration_hook_t iteration_hook_
mrpt::system::CTimeLogger profiler_ = {false, "mp2p_icp::ICP"}

Protected Static Functions

static void save_log_file(const LogRecord &log, const Parameters &p)
struct ICP_State

Public Functions

inline ICP_State(const metric_map_t &pcsGlobal, const metric_map_t &pcsLocal)

Public Members

const metric_map_t &pcGlobal
const metric_map_t &pcLocal
Pairings currentPairings
OptimalTF_Result currentSolution
uint32_t currentIteration = 0
LogRecord *log = nullptr
struct IterationHook_Input

Public Functions

IterationHook_Input() = default

Public Members

const metric_map_t *pcGlobal = nullptr
const metric_map_t *pcLocal = nullptr
const Pairings *currentPairings = nullptr
const OptimalTF_Result *currentSolution = nullptr
uint32_t currentIteration = 0
struct IterationHook_Output

Public Functions

IterationHook_Output() = default

Public Members

bool request_stop = false
struct QualityEvaluatorEntry

Public Functions

inline QualityEvaluatorEntry(mp2p_icp::QualityEvaluator::Ptr o, double w)

Public Members

mp2p_icp::QualityEvaluator::Ptr obj
double relativeWeight = 1.0