brute.cpp
Go to the documentation of this file.00001 //---------------------------------------------------------------------- 00002 // File: brute.cpp 00003 // Programmer: Sunil Arya and David Mount 00004 // Description: Brute-force nearest neighbors 00005 // Last modified: 05/03/05 (Version 1.1) 00006 //---------------------------------------------------------------------- 00007 // Copyright (c) 1997-2005 University of Maryland and Sunil Arya and 00008 // David Mount. All Rights Reserved. 00009 // 00010 // This software and related documentation is part of the Approximate 00011 // Nearest Neighbor Library (ANN). This software is provided under 00012 // the provisions of the Lesser GNU Public License (LGPL). See the 00013 // file ../ReadMe.txt for further information. 00014 // 00015 // The University of Maryland (U.M.) and the authors make no 00016 // representations about the suitability or fitness of this software for 00017 // any purpose. It is provided "as is" without express or implied 00018 // warranty. 00019 //---------------------------------------------------------------------- 00020 // History: 00021 // Revision 0.1 03/04/98 00022 // Initial release 00023 // Revision 1.1 05/03/05 00024 // Added fixed-radius kNN search 00025 //---------------------------------------------------------------------- 00026 00027 #include <stereo_wall_detection/ANN/ANNx.h> // all ANN includes 00028 #include "pr_queue_k.h" // k element priority queue 00029 00030 //---------------------------------------------------------------------- 00031 // Brute-force search simply stores a pointer to the list of 00032 // data points and searches linearly for the nearest neighbor. 00033 // The k nearest neighbors are stored in a k-element priority 00034 // queue (which is implemented in a pretty dumb way as well). 00035 // 00036 // If ANN_ALLOW_SELF_MATCH is ANNfalse then data points at distance 00037 // zero are not considered. 00038 // 00039 // Note that the error bound eps is passed in, but it is ignored. 00040 // These routines compute exact nearest neighbors (which is needed 00041 // for validation purposes in ann_test.cpp). 00042 //---------------------------------------------------------------------- 00043 00044 ANNbruteForce::ANNbruteForce( // constructor from point array 00045 ANNpointArray pa, // point array 00046 int n, // number of points 00047 int dd) // dimension 00048 { 00049 dim = dd; n_pts = n; pts = pa; 00050 } 00051 00052 ANNbruteForce::~ANNbruteForce() { } // destructor (empty) 00053 00054 void ANNbruteForce::annkSearch( // approx k near neighbor search 00055 ANNpoint q, // query point 00056 int k, // number of near neighbors to return 00057 ANNidxArray nn_idx, // nearest neighbor indices (returned) 00058 ANNdistArray dd, // dist to near neighbors (returned) 00059 double eps) // error bound (ignored) 00060 { 00061 ANNmin_k mk(k); // construct a k-limited priority queue 00062 int i; 00063 00064 if (k > n_pts) { // too many near neighbors? 00065 annError("Requesting more near neighbors than data points", ANNabort); 00066 } 00067 // run every point through queue 00068 for (i = 0; i < n_pts; i++) { 00069 // compute distance to point 00070 ANNdist sqDist = annDist(dim, pts[i], q); 00071 if (ANN_ALLOW_SELF_MATCH || sqDist != 0) 00072 mk.insert(sqDist, i); 00073 } 00074 for (i = 0; i < k; i++) { // extract the k closest points 00075 dd[i] = mk.ith_smallest_key(i); 00076 nn_idx[i] = mk.ith_smallest_info(i); 00077 } 00078 } 00079 00080 int ANNbruteForce::annkFRSearch( // approx fixed-radius kNN search 00081 ANNpoint q, // query point 00082 ANNdist sqRad, // squared radius 00083 int k, // number of near neighbors to return 00084 ANNidxArray nn_idx, // nearest neighbor array (returned) 00085 ANNdistArray dd, // dist to near neighbors (returned) 00086 double eps) // error bound 00087 { 00088 ANNmin_k mk(k); // construct a k-limited priority queue 00089 int i; 00090 int pts_in_range = 0; // number of points in query range 00091 // run every point through queue 00092 for (i = 0; i < n_pts; i++) { 00093 // compute distance to point 00094 ANNdist sqDist = annDist(dim, pts[i], q); 00095 if (sqDist <= sqRad && // within radius bound 00096 (ANN_ALLOW_SELF_MATCH || sqDist != 0)) { // ...and no self match 00097 mk.insert(sqDist, i); 00098 pts_in_range++; 00099 } 00100 } 00101 for (i = 0; i < k; i++) { // extract the k closest points 00102 if (dd != NULL) 00103 dd[i] = mk.ith_smallest_key(i); 00104 if (nn_idx != NULL) 00105 nn_idx[i] = mk.ith_smallest_info(i); 00106 } 00107 00108 return pts_in_range; 00109 }