OpenCV  3.4.14
Open Source Computer Vision
Public Types | Public Member Functions | List of all members
cv::flann::GenericIndex< Distance > Class Template Reference

The FLANN nearest neighbor index class. This class is templated with the type of elements for which the index is built. More...

#include <opencv2/flann.hpp>

Public Types

typedef Distance::ResultType DistanceType
 
typedef Distance::ElementType ElementType
 

Public Member Functions

 GenericIndex (const Mat &features, const ::cvflann::IndexParams &params, Distance distance=Distance())
 Constructs a nearest neighbor search index for a given dataset. More...
 
 ~GenericIndex ()
 
const ::cvflann::IndexParams * getIndexParameters ()
 
::cvflann::IndexParams getParameters ()
 
void knnSearch (const std::vector< ElementType > &query, std::vector< int > &indices, std::vector< DistanceType > &dists, int knn, const ::cvflann::SearchParams &params)
 Performs a K-nearest neighbor search for a given query point using the index. More...
 
void knnSearch (const Mat &queries, Mat &indices, Mat &dists, int knn, const ::cvflann::SearchParams &params)
 
int radiusSearch (const std::vector< ElementType > &query, std::vector< int > &indices, std::vector< DistanceType > &dists, DistanceType radius, const ::cvflann::SearchParams &params)
 Performs a radius nearest neighbor search for a given query point using the index. More...
 
int radiusSearch (const Mat &query, Mat &indices, Mat &dists, DistanceType radius, const ::cvflann::SearchParams &params)
 
void save (String filename)
 
int size () const
 
int veclen () const
 

Detailed Description

template<typename Distance>
class cv::flann::GenericIndex< Distance >

The FLANN nearest neighbor index class. This class is templated with the type of elements for which the index is built.

Distance functor specifies the metric to be used to calculate the distance between two points. There are several Distance functors that are readily available:

cv::cvflann::L2_Simple - Squared Euclidean distance functor. This is the simpler, unrolled version. This is preferable for very low dimensionality data (eg 3D points)

cv::flann::L2 - Squared Euclidean distance functor, optimized version.

cv::flann::L1 - Manhattan distance functor, optimized version.

cv::flann::MinkowskiDistance - The Minkowsky distance functor. This is highly optimised with loop unrolling. The computation of squared root at the end is omitted for efficiency.

cv::flann::MaxDistance - The max distance functor. It computes the maximum distance between two vectors. This distance is not a valid kdtree distance, it's not dimensionwise additive.

cv::flann::HammingLUT - Hamming distance functor. It counts the bit differences between two strings using a lookup table implementation.

cv::flann::Hamming - Hamming distance functor. Population count is performed using library calls, if available. Lookup table implementation is used as a fallback.

cv::flann::Hamming2 - Hamming distance functor. Population count is implemented in 12 arithmetic operations (one of which is multiplication).

cv::flann::DNAmmingLUT - Adaptation of the Hamming distance functor to DNA comparison. As the four bases A, C, G, T of the DNA (or A, G, C, U for RNA) can be coded on 2 bits, it counts the bits pairs differences between two sequences using a lookup table implementation.

cv::flann::DNAmming2 - Adaptation of the Hamming distance functor to DNA comparison. Bases differences count are vectorised thanks to arithmetic operations using standard registers (AVX2 and AVX-512 should come in a near future).

cv::flann::HistIntersectionDistance - The histogram intersection distance functor.

cv::flann::HellingerDistance - The Hellinger distance functor.

cv::flann::ChiSquareDistance - The chi-square distance functor.

cv::flann::KL_Divergence - The Kullback-Leibler divergence functor.

Although the provided implementations cover a vast range of cases, it is also possible to use a custom implementation. The distance functor is a class whose operator() computes the distance between two features. If the distance is also a kd-tree compatible distance, it should also provide an accum_dist() method that computes the distance between individual feature dimensions.

In addition to operator() and accum_dist(), a distance functor should also define the ElementType and the ResultType as the types of the elements it operates on and the type of the result it computes. If a distance functor can be used as a kd-tree distance (meaning that the full distance between a pair of features can be accumulated from the partial distances between the individual dimensions) a typedef is_kdtree_distance should be present inside the distance functor. If the distance is not a kd-tree distance, but it's a distance in a vector space (the individual dimensions of the elements it operates on can be accessed independently) a typedef is_vector_space_distance should be defined inside the functor. If neither typedef is defined, the distance is assumed to be a metric distance and will only be used with indexes operating on generic metric distances.

Member Typedef Documentation

◆ DistanceType

template<typename Distance >
typedef Distance::ResultType cv::flann::GenericIndex< Distance >::DistanceType

◆ ElementType

template<typename Distance >
typedef Distance::ElementType cv::flann::GenericIndex< Distance >::ElementType

Constructor & Destructor Documentation

◆ GenericIndex()

template<typename Distance >
cv::flann::GenericIndex< Distance >::GenericIndex ( const Mat features,
const ::cvflann::IndexParams &  params,
Distance  distance = Distance() 
)

Constructs a nearest neighbor search index for a given dataset.

Parameters
featuresMatrix of containing the features(points) to index. The size of the matrix is num_features x feature_dimensionality and the data type of the elements in the matrix must coincide with the type of the index.
paramsStructure containing the index parameters. The type of index that will be constructed depends on the type of this parameter. See the description.
distanceThe method constructs a fast search structure from a set of features using the specified algorithm with specified parameters, as defined by params. params is a reference to one of the following class IndexParams descendants:
  • LinearIndexParams When passing an object of this type, the index will perform a linear, brute-force search. :
    struct LinearIndexParams : public IndexParams
    {
    };
  • KDTreeIndexParams When passing an object of this type the index constructed will consist of a set of randomized kd-trees which will be searched in parallel. :
    struct KDTreeIndexParams : public IndexParams
    {
    KDTreeIndexParams( int trees = 4 );
    };
  • HierarchicalClusteringIndexParams When passing an object of this type the index constructed will be a hierarchical tree of clusters, dividing each set of points into n clusters whose centers are picked among the points without further refinement of their position. This algorithm fits both floating, integer and binary vectors. :
    struct HierarchicalClusteringIndexParams : public IndexParams
    {
    HierarchicalClusteringIndexParams(
    int branching = 32,
    flann_centers_init_t centers_init = CENTERS_RANDOM,
    int trees = 4,
    int leaf_size = 100);
    };
  • KMeansIndexParams When passing an object of this type the index constructed will be a hierarchical k-means tree (one tree by default), dividing each set of points into n clusters whose barycenters are refined iteratively. Note that this algorithm has been extended to the support of binary vectors as an alternative to LSH when knn search speed is the criterium. It will also outperform LSH when processing directly (i.e. without the use of MCA/PCA) datasets whose points share mostly the same values for most of the dimensions. It is recommended to set more than one tree with binary data. :
    struct KMeansIndexParams : public IndexParams
    {
    KMeansIndexParams(
    int branching = 32,
    int iterations = 11,
    flann_centers_init_t centers_init = CENTERS_RANDOM,
    float cb_index = 0.2 );
    KMeansIndexParams(
    int branching,
    int iterations,
    flann_centers_init_t centers_init,
    float cb_index,
    int trees );
    };
  • CompositeIndexParams When using a parameters object of this type the index created combines the randomized kd-trees and the hierarchical k-means tree. :
    struct CompositeIndexParams : public IndexParams
    {
    CompositeIndexParams(
    int trees = 4,
    int branching = 32,
    int iterations = 11,
    flann_centers_init_t centers_init = CENTERS_RANDOM,
    float cb_index = 0.2 );
    };
  • LshIndexParams When using a parameters object of this type the index created uses multi-probe LSH (by Multi-Probe LSH: Efficient Indexing for High-Dimensional Similarity Search by Qin Lv, William Josephson, Zhe Wang, Moses Charikar, Kai Li., Proceedings of the 33rd International Conference on Very Large Data Bases (VLDB). Vienna, Austria. September 2007). This algorithm is designed for binary vectors. :
    struct LshIndexParams : public IndexParams
    {
    LshIndexParams(
    unsigned int table_number,
    unsigned int key_size,
    unsigned int multi_probe_level );
    };
  • AutotunedIndexParams When passing an object of this type the index created is automatically tuned to offer the best performance, by choosing the optimal index type (randomized kd-trees, hierarchical kmeans, linear) and parameters for the dataset provided. :
    struct AutotunedIndexParams : public IndexParams
    {
    AutotunedIndexParams(
    float target_precision = 0.9,
    float build_weight = 0.01,
    float memory_weight = 0,
    float sample_fraction = 0.1 );
    };
  • SavedIndexParams This object type is used for loading a previously saved index from the disk. :
    struct SavedIndexParams : public IndexParams
    {
    SavedIndexParams( String filename );
    };

◆ ~GenericIndex()

template<typename Distance >
cv::flann::GenericIndex< Distance >::~GenericIndex ( )

Member Function Documentation

◆ getIndexParameters()

template<typename Distance >
const ::cvflann::IndexParams* cv::flann::GenericIndex< Distance >::getIndexParameters ( )
inline

◆ getParameters()

template<typename Distance >
::cvflann::IndexParams cv::flann::GenericIndex< Distance >::getParameters ( )
inline

◆ knnSearch() [1/2]

template<typename Distance >
void cv::flann::GenericIndex< Distance >::knnSearch ( const std::vector< ElementType > &  query,
std::vector< int > &  indices,
std::vector< DistanceType > &  dists,
int  knn,
const ::cvflann::SearchParams &  params 
)

Performs a K-nearest neighbor search for a given query point using the index.

Parameters
queryThe query point
indicesVector that will contain the indices of the K-nearest neighbors found. It must have at least knn size.
distsVector that will contain the distances to the K-nearest neighbors found. It must have at least knn size.
knnNumber of nearest neighbors to search for.
paramsSearchParams

◆ knnSearch() [2/2]

template<typename Distance >
void cv::flann::GenericIndex< Distance >::knnSearch ( const Mat queries,
Mat indices,
Mat dists,
int  knn,
const ::cvflann::SearchParams &  params 
)

◆ radiusSearch() [1/2]

template<typename Distance >
int cv::flann::GenericIndex< Distance >::radiusSearch ( const std::vector< ElementType > &  query,
std::vector< int > &  indices,
std::vector< DistanceType > &  dists,
DistanceType  radius,
const ::cvflann::SearchParams &  params 
)

Performs a radius nearest neighbor search for a given query point using the index.

Parameters
queryThe query point.
indicesVector that will contain the indices of the nearest neighbors found.
distsVector that will contain the distances to the nearest neighbors found. It has the same number of elements as indices.
radiusThe search radius.
paramsSearchParams

This function returns the number of nearest neighbors found.

◆ radiusSearch() [2/2]

template<typename Distance >
int cv::flann::GenericIndex< Distance >::radiusSearch ( const Mat query,
Mat indices,
Mat dists,
DistanceType  radius,
const ::cvflann::SearchParams &  params 
)

◆ save()

template<typename Distance >
void cv::flann::GenericIndex< Distance >::save ( String  filename)
inline

◆ size()

template<typename Distance >
int cv::flann::GenericIndex< Distance >::size ( ) const
inline

◆ veclen()

template<typename Distance >
int cv::flann::GenericIndex< Distance >::veclen ( ) const
inline

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