1 #ifndef KDTREE_H 2 #define KDTREE_H 3 4 #include "precomp.hpp" 5 6 namespace cv 7 { 8 namespace ml 9 { 10 11 /*! 12 Fast Nearest Neighbor Search Class. 13 14 The class implements D. Lowe BBF (Best-Bin-First) algorithm for the last 15 approximate (or accurate) nearest neighbor search in multi-dimensional spaces. 16 17 First, a set of vectors is passed to KDTree::KDTree() constructor 18 or KDTree::build() method, where it is reordered. 19 20 Then arbitrary vectors can be passed to KDTree::findNearest() methods, which 21 find the K nearest neighbors among the vectors from the initial set. 22 The user can balance between the speed and accuracy of the search by varying Emax 23 parameter, which is the number of leaves that the algorithm checks. 24 The larger parameter values yield more accurate results at the expense of lower processing speed. 25 26 \code 27 KDTree T(points, false); 28 const int K = 3, Emax = INT_MAX; 29 int idx[K]; 30 float dist[K]; 31 T.findNearest(query_vec, K, Emax, idx, 0, dist); 32 CV_Assert(dist[0] <= dist[1] && dist[1] <= dist[2]); 33 \endcode 34 */ 35 class CV_EXPORTS_W KDTree 36 { 37 public: 38 /*! 39 The node of the search tree. 40 */ 41 struct Node 42 { Nodecv::ml::KDTree::Node43 Node() : idx(-1), left(-1), right(-1), boundary(0.f) {} Nodecv::ml::KDTree::Node44 Node(int _idx, int _left, int _right, float _boundary) 45 : idx(_idx), left(_left), right(_right), boundary(_boundary) {} 46 47 //! split dimension; >=0 for nodes (dim), < 0 for leaves (index of the point) 48 int idx; 49 //! node indices of the left and the right branches 50 int left, right; 51 //! go to the left if query_vec[node.idx]<=node.boundary, otherwise go to the right 52 float boundary; 53 }; 54 55 //! the default constructor 56 CV_WRAP KDTree(); 57 //! the full constructor that builds the search tree 58 CV_WRAP KDTree(InputArray points, bool copyAndReorderPoints = false); 59 //! the full constructor that builds the search tree 60 CV_WRAP KDTree(InputArray points, InputArray _labels, 61 bool copyAndReorderPoints = false); 62 //! builds the search tree 63 CV_WRAP void build(InputArray points, bool copyAndReorderPoints = false); 64 //! builds the search tree 65 CV_WRAP void build(InputArray points, InputArray labels, 66 bool copyAndReorderPoints = false); 67 //! finds the K nearest neighbors of "vec" while looking at Emax (at most) leaves 68 CV_WRAP int findNearest(InputArray vec, int K, int Emax, 69 OutputArray neighborsIdx, 70 OutputArray neighbors = noArray(), 71 OutputArray dist = noArray(), 72 OutputArray labels = noArray()) const; 73 //! finds all the points from the initial set that belong to the specified box 74 CV_WRAP void findOrthoRange(InputArray minBounds, 75 InputArray maxBounds, 76 OutputArray neighborsIdx, 77 OutputArray neighbors = noArray(), 78 OutputArray labels = noArray()) const; 79 //! returns vectors with the specified indices 80 CV_WRAP void getPoints(InputArray idx, OutputArray pts, 81 OutputArray labels = noArray()) const; 82 //! return a vector with the specified index 83 const float* getPoint(int ptidx, int* label = 0) const; 84 //! returns the search space dimensionality 85 CV_WRAP int dims() const; 86 87 std::vector<Node> nodes; //!< all the tree nodes 88 CV_PROP Mat points; //!< all the points. It can be a reordered copy of the input vector set or the original vector set. 89 CV_PROP std::vector<int> labels; //!< the parallel array of labels. 90 CV_PROP int maxDepth; //!< maximum depth of the search tree. Do not modify it 91 CV_PROP_RW int normType; //!< type of the distance (cv::NORM_L1 or cv::NORM_L2) used for search. Initially set to cv::NORM_L2, but you can modify it 92 }; 93 94 } 95 } 96 97 #endif 98