Frank Neuhaus > Code Snippets

Code Snippets

All code snippets typically require C++11.

Simple 3d Plane RANSAC

This code robustly fits a 3-dimensional plane through a given point cloud using the RANSAC algorithm. The implementation is rather straightforward, yet it can easily be adapted to other fitting problems.

Download

Example usage...

#include <iostream>
#include <vector>
#include <random>
#include <Eigen/Core>
#include "SimpleRansac.h"
#include "PlaneModel.h"
 
int main()
{
    std::random_device dev;
    std::mt19937 engine{dev()};
    std::uniform_real_distribution<double> dis{-1.0,1.0};
    auto draw=[&]() { return dis(engine); };
 
    std::vector<Eigen::Vector3d> points;
    // add some outliers (1000 randomly sampled points with coordinates in [-1,1])
    for (int i=0;i<1000;i++)
        points.push_back(Eigen::Vector3d(draw(),draw(),draw()));
    // add some inliers (200 randomly sampled points on a plane through point p, spanned by the vectors v1, v2)
    const Eigen::Vector3d p(2.0,2.0,0.0);
    const Eigen::Vector3d v1(0.3,0.7,0.2);
    const Eigen::Vector3d v2(-0.3,0.3,0.0);
    for (int i=0;i<200;i++)
        points.push_back(p+draw()*v1+draw()*v2);
 
    // compute ground truth plane
    const Eigen::Vector3d n=v1.cross(v2).normalized();
    const double d=n.transpose()*p;
    std::cout << "Ground truth plane: n=" << n.transpose() << " d=" << d << std::endl;
    
    // run ransac
    // we do 300 iterations
    // points need to be closer than 0.05 in order to be considered an inlier
    const PlaneModel m=ransac<PlaneModel>(points,0.05,300);
    std::cout << "Ransac computed plane: n=" << m.n.transpose() << " d=" << m.d << std::endl;
}

</div>

Expected output of this program:

$ g++ -std=c++11 -I/usr/include/eigen3 main.cpp && ./a.out
 
Ground truth plane: n=-0.19245 -0.19245  0.96225 d=-0.7698
Ransac computed plane: n= 0.193933  0.190752 -0.962291 d=0.769292

Simple 2d Marching Squares

A simple implementation of the marching squares algorithm in 2d.
Depends on the Eigen Library.
The function to evaluate the iso surface can be provided by the user.
The algorithm yields the triangles of the ‘inside’ part, as well as a number of line segments forming the border.

Download

Simple Numerical Differentiation with Eigen

Download

https://github.com/cfneuhaus/snippets/blob/master/NumJac.h

Usage example:

#include <iostream>
#include <Eigen/Core>
#include "NumJac.h"

Eigen::Vector2d f(const Eigen::Vector3d& x)
{
    Eigen::Vector2d ret;
    ret << x(0)*x(0)+5.0*x(1)+x(2),
           x(0)+7.0*x(2);
    return ret;
}

int main()
{
    const Eigen::Matrix<double,2,3> jac=numJacobian(f,Eigen::Vector3d(1,2,3));
    std::cout << jac << std::endl;
}

Find In File/Source Scripts for Linux Bash

Find in Files (fif):

#!/bin/sh
 
find . \( -name '*' \) -print0 -follow 2>/dev/null | grep -z -Z -v -e '/CVS/' -e '/SCCS/' -e '/\.svn/' -e '/_darcs/' | xargs -0 egrep --color=auto -H -n -s -i -I -e $*

Find in Source (fis):

#!/bin/sh

find . \( -name '*.py' -o -name '*.h' -o -name '*.hxx' -o -name '*.hpp' -o -name '*.hh' -o -name '*.h++' -o -name '*.H' -o -name '*.tlh' -o -name '*.cpp' -o -name '*.cc' -o -name '*.C' -o -name '*.c++' -o -name '*.cxx' -o -name '*.ocl' -o -name '*.inl' -o -name '*.idl' -o -name '*.c' -o -name '*.m' -o -name '*.mm' -o -name '*.M' \) -print -follow 2>/dev/null | grep -v -e '/CVS/' -e '/SCCS/' -e '/\.svn/' -e '/_darcs/' | sed "s/ /\\\ /g" | xargs egrep --color=auto -H -n -s -i -e "$1"