Go to the documentation of this file.00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022
00023
00024
00025
00026
00027
00028
00029
00030
00031 #ifndef _OPENCV_COMPOSITETREE_H_
00032 #define _OPENCV_COMPOSITETREE_H_
00033
00034 #include "opencv2/flann/general.h"
00035 #include "opencv2/flann/nn_index.h"
00036
00037 namespace cvflann
00038 {
00039
00040
00041 struct CompositeIndexParams : public IndexParams {
00042 CompositeIndexParams(int trees_ = 4, int branching_ = 32, int iterations_ = 11,
00043 flann_centers_init_t centers_init_ = FLANN_CENTERS_RANDOM, float cb_index_ = 0.2 ) :
00044 IndexParams(FLANN_INDEX_COMPOSITE),
00045 trees(trees_),
00046 branching(branching_),
00047 iterations(iterations_),
00048 centers_init(centers_init_),
00049 cb_index(cb_index_) {};
00050
00051 int trees;
00052 int branching;
00053 int iterations;
00054 flann_centers_init_t centers_init;
00055 float cb_index;
00056
00057 void print() const
00058 {
00059 logger().info("Index type: %d\n",(int)algorithm);
00060 logger().info("Trees: %d\n", trees);
00061 logger().info("Branching: %d\n", branching);
00062 logger().info("Iterations: %d\n", iterations);
00063 logger().info("Centres initialisation: %d\n", centers_init);
00064 logger().info("Cluster boundary weight: %g\n", cb_index);
00065 }
00066 };
00067
00068
00069
00070 template <typename ELEM_TYPE, typename DIST_TYPE = typename DistType<ELEM_TYPE>::type >
00071 class CompositeIndex : public NNIndex<ELEM_TYPE>
00072 {
00073 KMeansIndex<ELEM_TYPE, DIST_TYPE>* kmeans;
00074 KDTreeIndex<ELEM_TYPE, DIST_TYPE>* kdtree;
00075
00076 const Matrix<ELEM_TYPE> dataset;
00077
00078 const IndexParams& index_params;
00079
00080 CompositeIndex& operator=(const CompositeIndex&);
00081 CompositeIndex(const CompositeIndex&);
00082 public:
00083
00084 CompositeIndex(const Matrix<ELEM_TYPE>& inputData, const CompositeIndexParams& params = CompositeIndexParams() ) :
00085 dataset(inputData), index_params(params)
00086 {
00087 KDTreeIndexParams kdtree_params(params.trees);
00088 KMeansIndexParams kmeans_params(params.branching, params.iterations, params.centers_init, params.cb_index);
00089
00090 kdtree = new KDTreeIndex<ELEM_TYPE, DIST_TYPE>(inputData,kdtree_params);
00091 kmeans = new KMeansIndex<ELEM_TYPE, DIST_TYPE>(inputData,kmeans_params);
00092
00093 }
00094
00095 virtual ~CompositeIndex()
00096 {
00097 delete kdtree;
00098 delete kmeans;
00099 }
00100
00101
00102 flann_algorithm_t getType() const
00103 {
00104 return FLANN_INDEX_COMPOSITE;
00105 }
00106
00107
00108 size_t size() const
00109 {
00110 return dataset.rows;
00111 }
00112
00113 size_t veclen() const
00114 {
00115 return dataset.cols;
00116 }
00117
00118
00119 int usedMemory() const
00120 {
00121 return kmeans->usedMemory()+kdtree->usedMemory();
00122 }
00123
00124 void buildIndex()
00125 {
00126 logger().info("Building kmeans tree...\n");
00127 kmeans->buildIndex();
00128 logger().info("Building kdtree tree...\n");
00129 kdtree->buildIndex();
00130 }
00131
00132
00133 void saveIndex(FILE* stream)
00134 {
00135 kmeans->saveIndex(stream);
00136 kdtree->saveIndex(stream);
00137 }
00138
00139
00140 void loadIndex(FILE* stream)
00141 {
00142 kmeans->loadIndex(stream);
00143 kdtree->loadIndex(stream);
00144 }
00145
00146 void findNeighbors(ResultSet<ELEM_TYPE>& result, const ELEM_TYPE* vec, const SearchParams& searchParams)
00147 {
00148 kmeans->findNeighbors(result,vec,searchParams);
00149 kdtree->findNeighbors(result,vec,searchParams);
00150 }
00151
00152 const IndexParams* getParameters() const
00153 {
00154 return &index_params;
00155 }
00156
00157
00158 };
00159
00160 }
00161
00162 #endif //_OPENCV_COMPOSITETREE_H_