File: mbl_cluster_tree.txx

package info (click to toggle)
vxl 1.17.0.dfsg-1
  • links: PTS, VCS
  • area: main
  • in suites: jessie, jessie-kfreebsd
  • size: 153,280 kB
  • ctags: 105,123
  • sloc: cpp: 747,420; ansic: 209,130; fortran: 34,230; lisp: 14,915; sh: 6,187; python: 5,856; makefile: 340; perl: 294; xml: 160
file content (308 lines) | stat: -rw-r--r-- 8,488 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
// This is mul/mbl/mbl_cluster_tree.txx
#ifndef mbl_cluster_txx_
#define mbl_cluster_txx_
#ifdef VCL_NEEDS_PRAGMA_INTERFACE
#pragma interface
#endif
//:
// \file
// \brief  Record trees of clusters of data, for faster neighbour finding
// \author Tim Cootes

#include "mbl_cluster_tree.h"
#include <vcl_iostream.h>
#include <vcl_cassert.h>
#include <vsl/vsl_binary_io.h>
#include <vsl/vsl_vector_io.h>

//: Default constructor
template<class T, class D>
mbl_cluster_tree<T,D>::mbl_cluster_tree()
  : data_(0)
{
}

//: Empty clusters
template<class T, class D>
void mbl_cluster_tree<T,D>::empty()
{
  for (unsigned i=0;i<cluster_.size();++i)
  {
    cluster_[i].empty();
    parent_[i].resize(0);
  }
}

//: Define number of levels and max radius of clusters at each level
template<class T, class D>
void mbl_cluster_tree<T,D>::set_max_r(const vcl_vector<double>& r)
{
  empty();
  unsigned nL = r.size();
  cluster_.resize(nL);
  for (unsigned i=0;i<nL;++i) cluster_[i].set_max_r(r[i]);
  parent_.resize( nL);
  for (unsigned i=0;i<nL;++i) parent_[i].resize(0);

  cluster_[0].set_data(data_);
  for (unsigned i=1;i<nL;++i)
    cluster_[i].set_data(cluster_[i-1].p());
}


//: Define external data array (pointer retained)
//  Empty existing clusters, then process every element of data
//  to create clusters, by calling add_object()
template<class T, class D>
void mbl_cluster_tree<T,D>::set_data(const vcl_vector<T>& data)
{
  empty();
  data_ = data;
  parent_[0].resize(data.size());
  unsigned n = data.size();
  for (unsigned i=0;i<n;++i)  add_object(i);
}

//: Add an extra element to data()
template<class T, class D>
void mbl_cluster_tree<T,D>::push_back(const T& t)
{
  data_.push_back(t);
  parent_[0].push_back(0);  // Create space for parent
  add_object(data().size()-1);
}

//: Return index of nearest object in data() to t
//  Nearest object in data() to t is given by data()[nearest(t,d)];
//  The distance to the point is d
template<class T, class D>
unsigned mbl_cluster_tree<T,D>::nearest(const T& t, double& d) const
{
  assert(data().size()>0);
  assert(cluster_.size()>0);

  if (cluster_.size()==1) return cluster_[0].nearest(t,d);

  // Perform hierarchical search
  // Find possible clusters at top level
  unsigned L=cluster_.size()-1;
  vcl_vector<unsigned> near_c0, near_c1;
  double max_d;
  cluster_[L].nearest_clusters(t,max_d,near_c1);

  cluster_[L].in_clusters(near_c1,near_c0);  // Find objects in next L
  L--;
  while (L!=0)
  {
    cluster_[L].nearest_clusters(t,max_d,near_c0,near_c1);
    cluster_[L].in_clusters(near_c1,near_c0);   // Find objects in next L
    --L;
  }
  return cluster_[0].nearest(t,d,near_c0);
}

//: Append new object with index i and assign to clusters
//  Assumes that new object data()[i] is available.
//  Deduce which cluster it belongs to and add it.
//  Create new cluster if further than max_r() from any.
template<class T, class D>
void mbl_cluster_tree<T,D>::add_object(unsigned new_i)
{
  assert(new_i<data().size());

  if (cluster_.size()==1) { cluster_[0].add_object(new_i); return; }

  const T& t = data()[new_i];

  unsigned Lhi = cluster_.size()-1;

  // Find any clusters at top level which could hold t
  vcl_vector<unsigned> c_list0;
  vcl_vector<unsigned>  nearest_c(Lhi+1);
  vcl_vector<double>  nearest_d(Lhi+1);

  if (cluster_[Lhi].clusters_within_max_r(t,c_list0,nearest_c[Lhi],
                                                   nearest_d[Lhi])==0)
  {
    // No clusters at the top level can include t, so create new
    // cluster at each level.
    unsigned cL0=cluster_[0].create_cluster(new_i);
    parent_[0][new_i]=cL0;
    for (unsigned L=1;L<=Lhi;++L)
    {
      unsigned cL1=cluster_[L].create_cluster(cL0);
      parent_[L].push_back(cL1);
      cL0=cL1;  // Record position
    }
    return;
  }

  for (unsigned L0=Lhi;L0!=0;--L0)
  {
    unsigned L = L0 - 1;
    vcl_vector<unsigned> c_list1=c_list0;

     // Generate list of elements in each cluster to process at next level
    cluster_[L0].in_clusters(c_list1,c_list0);

    // Find clusters at level L which could contain t
    unsigned nc=cluster_[L].clusters_within_max_r(t,c_list0,c_list1,
                                       nearest_c[L],nearest_d[L]);

    if (nc==0)
    {
      // No clusters at this level can include t,
      // so create new cluster at levels 0..L
      unsigned cL=cluster_[0].create_cluster(new_i);
      parent_[0][new_i]=cL;
      if (parent_.size()>1)
        parent_[1].push_back(0);  // Make space for new one

      for (unsigned L1=1;L1<=L;++L1)
      {
        unsigned cL1=cluster_[L1].create_cluster(cL);
        if (L1<Lhi)
          parent_[L1+1].push_back(0);  // Create space for record of new
        parent_[L1][cL]=cL1;
        cL=cL1;  // Record index
      }

      // cL is cluster containing new object at level L
      // Assign to nearest cluster in level above (c)
      unsigned c=nearest_c[L0];
      cluster_[L0].assign_to_cluster(cL,c,nearest_d[L0]);
      parent_[L0][cL]=c;  // Record parent

        // Track back through ancestors of c, updating radii
      for (unsigned L1=L0+1;L1<=Lhi;++L1)
      {
        c = parent_[L1][c];
        double d=D::d(t,cluster_[L1].p()[c]);
        if (d>cluster_[L1].r()[c])
          cluster_[L1].set_r(c,d);
      }
      return;
    }
    vcl_swap(c_list0,c_list1);  // Set c_list0 to current valid list
  }

  // If reached here, then t is in range of a cluster at every level
  unsigned c = nearest_c[0];
  cluster_[0].assign_to_cluster(new_i,c,nearest_d[0]);
  parent_[0][new_i]=c;
  for (unsigned L=1;L<=Lhi;++L)
  {
    // Track back through parents of c, updating radii
    c = parent_[L][c];
    double d=D::d(t,cluster_[L].p()[c]);
    if (d>cluster_[L].r()[c])
      cluster_[L].set_r(c,d);
  }
}

//: Print ancestry of every element
template<class T, class D>
void mbl_cluster_tree<T,D>::print_tree(vcl_ostream& os) const
{
  for (unsigned i=0;i<data().size();++i)
  {
    os << i;
    unsigned p=i;
    for (unsigned L=0;L<cluster_.size();++L)
    {
      p=parent_[L][p];
      os << " - "<<p;
    }
    os<<'\n';
  }
}

//: Print summary information
template<class T, class D>
void mbl_cluster_tree<T,D>::print_summary(vcl_ostream& os) const
{
  for (unsigned i=0;i<cluster_.size();++i)
  {
    os << "Level "<<i<<") max_r: "<<cluster_[i].max_r()
       << " n_clusters: "<<cluster_[i].p().size()<<'\n';
  }
}

template<class T, class D>
short mbl_cluster_tree<T,D>::version_no() const
{
    return 1;
}

template<class T, class D>
void mbl_cluster_tree<T,D>::b_write(vsl_b_ostream& bfs) const
{
  vsl_b_write(bfs,version_no());
  vsl_b_write(bfs,data_);
  vsl_b_write(bfs,parent_);
  // Write out clusters explicitly to avoid creating another template
  vsl_b_write(bfs,cluster_.size());
  for (unsigned L=0;L<cluster_.size();++L)
    vsl_b_write(bfs,cluster_[L]);
}

template<class T, class D>
void mbl_cluster_tree<T,D>::b_read(vsl_b_istream& bfs)
{
  short version;
  vsl_b_read(bfs,version);
  unsigned nc=0;
  switch (version)
  {
    case 1:
      vsl_b_read(bfs,data_);
      vsl_b_read(bfs,parent_);
      vsl_b_read(bfs,nc);
      cluster_.resize(nc);
      for (unsigned L=0;L<nc;++L)
        vsl_b_read(bfs,cluster_[L]);
   break;

  default:
    vcl_cerr << "mbl_cluster_tree<T,D>::b_read() "
      "Unexpected version number " << version << vcl_endl;
    bfs.is().clear(vcl_ios::badbit); // Set an unrecoverable IO error on stream
    return;
  }

  // Connect each cluster to data below
  cluster_[0].set_data(data_);
  for (unsigned L=1;L<cluster_.size();++L)
    cluster_[L].set_data(cluster_[L-1].p());
}

//: Binary file stream output operator for class reference
template<class T, class D>
void vsl_b_write(vsl_b_ostream& bfs, const mbl_cluster_tree<T,D>& c)
{
  c.b_write(bfs);
}

//: Binary file stream input operator for class reference
template<class T, class D>
void vsl_b_read(vsl_b_istream& bfs, mbl_cluster_tree<T,D>& c)
{
  c.b_read(bfs);
}

//: Stream output operator for class reference
template<class T, class D>
vcl_ostream& operator<<(vcl_ostream& os,const mbl_cluster_tree<T,D>& c)
{
  c.print_summary(os);
  return os;
}


#define MBL_CLUSTER_TREE_INSTANTIATE(T,D) \
template class mbl_cluster_tree<T,D >; \
template void vsl_b_write(vsl_b_ostream& bfs, const mbl_cluster_tree<T,D >& c); \
template void vsl_b_read(vsl_b_istream& bfs, mbl_cluster_tree<T,D >& c); \
template vcl_ostream& operator<<(vcl_ostream&os,const mbl_cluster_tree<T,D >& c)

#endif // mbl_cluster_txx_