snark Namespace Reference

Classes
struct	array_hash
struct	constant_speed
	point motion model with constant speed More...
class	gaussian_process
	gaussian process More...
class	kalman_filter
	generic kalman filter More...
class	pcap_reader
class	pin_screen
class	proprietary_reader
class	range_bearing_elevation
	polar point definition (range-bearing-elevation) More...
class	rotation_matrix
	rotation matrix (a quick and dirty class, to refactor at some point) More...
class	stdin_reader
	stdin reader More...
class	thin_reader
	reader for thinned velodyne data More...
class	timer
	simple timer More...
class	udp_reader
	udp reader More...
struct	velodyne_point
	processed velodyne point More...
class	velodyne_stream
	convert stream of raw velodyne data into velodyne points More...
class	velodyne_stream< comma::csv::input_stream< velodyne_point > >
	specialisation for csv input stream: in this case nothing to convert More...
class	voxel_grid
class	voxel_map

Functions
template<typename It , typename N , typename Tr >
std::map< comma::uint32, std::list< It > >	equivalence_classes (const It &begin, const It &end, comma::uint32 minId)
	partition elements of container
template<typename T , std::size_t N>
void	fft (boost::array< T, N > &data)
template<typename T >
void	fft (T *data, std::size_t size)
	Cooley-Tukey in-place fft based on Danielson-Lanczos algorithm.
template<typename T , std::size_t N>
boost::array< T, N >	fft (const boost::array< T, N > &data)
	Cooley-Tukey fft based on Danielson-Lanczos algorithm (convenience function)
template<typename It , typename Id , typename F >
Id	voted_tracking (It begin, It end, F GetPreviousId, Id vacantId)

Detailed Description

Author

Vsevolod Vlaskine

Vsevolod Vlaskine, Cedric Wohlleber

Cedric Wohlleber

vsevolod vlaskine

Function Documentation

void fft

(

boost::array< T, N > &

data

)

Cooley-Tukey in-place fft based on Danielson-Lanczos algorithm

see: Cooley-Tukey algorithm from Numerical Recipes in C++

see also: Vlodymyr Myrnyy, Simple and Efficient Fast Fourier Transform, Dr. Dobbs' Journal http://www.drdobbs.com/cpp/a-simple-and-efficient-fft-implementatio/199500857

Referenced by fft().

Id voted_tracking ( It  begin, It  end, F  GetPreviousId, Id  vacantId  )

inline

A common problem:

We have two subsequent velodyne scans partitioned.
Now we want to colour them. But the partitioning algorithms often
don't guarantee that the partition ids will be the same:

A tree might have partition id = 5 in partition #1, and 6 in partition #2,
although it is the same tree.

Therefore, in GUI, the tree will first be, say, blue and then suddenly
turn pink.

The following algorithm solves this problem for static and slow-moving objects by simple voting.

It works in one pass (i.e. ~O(N)) by the price of additional memory (as we need to store the previous id - e.g. previous colour).

See unit test for usage example