RobustKernel

base for all robust cost functions More...

Inheritance diagram for RobustKernel:

Public Member Functions
g_type	delta () const
	Returns the current delta (window size) value.
virtual void	robustify (g_type squaredError, Eigen::Vector3< g_type > &rho) const =0
	compute the scaling factor for a error: The error is e^T Omega e The output rho is rho[0]: The actual scaled error value rho[1]: First derivative of the scaling function rho[2]: Second derivative of the scaling function
virtual void	setDelta (g_type delta)
	set the window size of the error.

Protected Attributes
g_type	_delta = 1.0
	The window size parameter; errors above delta^2 are considered outliers.

Detailed Description

base for all robust cost functions

Note in all the implementations for the other cost functions the e in the funtions corresponds to the sqaured errors, i.e., the robustification functions gets passed the squared error.

e.g. the robustified least squares function is

chi^2 = sum_{e} rho( e^T Omega e )

Definition at line 19 of file robust_kernel.h.

Member Function Documentation

setDelta()

virtual void RobustKernel::setDelta ( g_type delta )

inlinevirtual

set the window size of the error.

A squared error above delta^2 is considered as outlier in the data.

Reimplemented in RobustKernelHuber.

Definition at line 43 of file robust_kernel.h.

References _delta, and delta().

---

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

• GraphOptimization/Headers/robust_kernel.h