icp.cpp 13 KB
Newer Older
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
//
//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
//  By downloading, copying, installing or using the software you agree to this license.
//  If you do not agree to this license, do not download, install,
//  copy or use the software.
//
//
//                          License Agreement
//                For Open Source Computer Vision Library
//
// Copyright (C) 2014, OpenCV Foundation, all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
//   * Redistribution's of source code must retain the above copyright notice,
//     this list of conditions and the following disclaimer.
//
//   * Redistribution's in binary form must reproduce the above copyright notice,
//     this list of conditions and the following disclaimer in the documentation
//     and/or other materials provided with the distribution.
//
//   * The name of the copyright holders may not be used to endorse or promote products
//     derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
// Author: Tolga Birdal <tbirdal AT gmail.com>

#include "precomp.hpp"

namespace cv
{
namespace ppf_match_3d
{
47
static void subtractColumns(Mat srcPC, Vec3d& mean)
48
{
Bence Magyar's avatar
Bence Magyar committed
49 50 51 52
  int height = srcPC.rows;

  for (int i=0; i<height; i++)
  {
53
    float *row = srcPC.ptr<float>(i);
54
    {
Bence Magyar's avatar
Bence Magyar committed
55 56 57
      row[0]-=(float)mean[0];
      row[1]-=(float)mean[1];
      row[2]-=(float)mean[2];
58
    }
Bence Magyar's avatar
Bence Magyar committed
59
  }
60 61 62
}

// as in PCA
63
static void computeMeanCols(Mat srcPC, Vec3d& mean)
64
{
Bence Magyar's avatar
Bence Magyar committed
65 66 67 68 69 70
  int height = srcPC.rows;

  double mean1=0, mean2 = 0, mean3 = 0;

  for (int i=0; i<height; i++)
  {
71
    const float *row = srcPC.ptr<float>(i);
72
    {
Bence Magyar's avatar
Bence Magyar committed
73 74 75
      mean1 += (double)row[0];
      mean2 += (double)row[1];
      mean3 += (double)row[2];
76
    }
Bence Magyar's avatar
Bence Magyar committed
77 78 79 80 81 82 83 84 85
  }

  mean1/=(double)height;
  mean2/=(double)height;
  mean3/=(double)height;

  mean[0] = mean1;
  mean[1] = mean2;
  mean[2] = mean3;
86 87 88
}

// as in PCA
89
/*static void subtractMeanFromColumns(Mat srcPC, Vec3d& mean)
90 91 92 93 94 95 96 97
{
    computeMeanCols(srcPC, mean);
    subtractColumns(srcPC, mean);
}*/

// compute the average distance to the origin
static double computeDistToOrigin(Mat srcPC)
{
Bence Magyar's avatar
Bence Magyar committed
98 99 100 101 102
  int height = srcPC.rows;
  double dist = 0;

  for (int i=0; i<height; i++)
  {
103
    const float *row = srcPC.ptr<float>(i);
Bence Magyar's avatar
Bence Magyar committed
104 105 106 107
    dist += sqrt(row[0]*row[0]+row[1]*row[1]+row[2]*row[2]);
  }

  return dist;
108 109 110 111 112
}

// From numerical receipes: Finds the median of an array
static float medianF(float arr[], int n)
{
Bence Magyar's avatar
Bence Magyar committed
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
  int low, high ;
  int median;
  int middle, ll, hh;

  low = 0 ;
  high = n-1 ;
  median = (low + high) >>1;
  for (;;)
  {
    if (high <= low) /* One element only */
      return arr[median] ;

    if (high == low + 1)
    {
      /* Two elements only */
      if (arr[low] > arr[high])
        std::swap(arr[low], arr[high]) ;
      return arr[median] ;
    }

    /* Find median of low, middle and high items; swap into position low */
    middle = (low + high) >>1;
    if (arr[middle] > arr[high])
      std::swap(arr[middle], arr[high]) ;
    if (arr[low] > arr[high])
      std::swap(arr[low], arr[high]) ;
    if (arr[middle] > arr[low])
      std::swap(arr[middle], arr[low]) ;

    /* Swap low item (now in position middle) into position (low+1) */
    std::swap(arr[middle], arr[low+1]) ;

    /* Nibble from each end towards middle, swapping items when stuck */
    ll = low + 1;
    hh = high;
148 149
    for (;;)
    {
Bence Magyar's avatar
Bence Magyar committed
150 151 152 153 154 155 156 157 158 159 160
      do
        ll++;
      while (arr[low] > arr[ll]) ;
      do
        hh--;
      while (arr[hh]  > arr[low]) ;

      if (hh < ll)
        break;

      std::swap(arr[ll], arr[hh]) ;
161
    }
Bence Magyar's avatar
Bence Magyar committed
162 163 164 165 166 167 168 169 170 171

    /* Swap middle item (in position low) back into correct position */
    std::swap(arr[low], arr[hh]) ;

    /* Re-set active partition */
    if (hh <= median)
      low = ll;
    if (hh >= median)
      high = hh - 1;
  }
172 173 174 175
}

static float getRejectionThreshold(float* r, int m, float outlierScale)
{
Bence Magyar's avatar
Bence Magyar committed
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
  float* t=(float*)calloc(m, sizeof(float));
  int i=0;
  float s=0, medR, threshold;

  memcpy(t, r, m*sizeof(float));
  medR=medianF(t, m);

  for (i=0; i<m; i++)
    t[i] = (float)fabs((double)r[i]-(double)medR);

  s = 1.48257968f * medianF(t, m);

  threshold = (outlierScale*s+medR);

  free(t);
  return threshold;
192 193 194
}

// Kok Lim Low's linearization
195
static void minimizePointToPlaneMetric(Mat Src, Mat Dst, Vec3d& rpy, Vec3d& t)
196
{
Bence Magyar's avatar
Bence Magyar committed
197 198 199
  //Mat sub = Dst - Src;
  Mat A = Mat(Src.rows, 6, CV_64F);
  Mat b = Mat(Src.rows, 1, CV_64F);
200
  Mat rpy_t;
Bence Magyar's avatar
Bence Magyar committed
201

202 203 204
#if defined _OPENMP
#pragma omp parallel for
#endif
Bence Magyar's avatar
Bence Magyar committed
205 206
  for (int i=0; i<Src.rows; i++)
  {
207 208 209 210 211 212 213 214
    const Vec3d srcPt(Src.ptr<double>(i));
    const Vec3d dstPt(Dst.ptr<double>(i));
    const Vec3d normals(Dst.ptr<double>(i) + 3);
    const Vec3d sub = dstPt - srcPt;
    const Vec3d axis = srcPt.cross(normals);

    *b.ptr<double>(i) = sub.dot(normals);
    hconcat(axis.reshape<1, 3>(), normals.reshape<1, 3>(), A.row(i));
Bence Magyar's avatar
Bence Magyar committed
215 216
  }

217 218 219
  cv::solve(A, b, rpy_t, DECOMP_SVD);
  rpy_t.rowRange(0, 3).copyTo(rpy);
  rpy_t.rowRange(3, 6).copyTo(t);
220 221
}

222
static void getTransformMat(Vec3d& euler, Vec3d& t, Matx44d& Pose)
223
{
224 225 226
  Matx33d R;
  eulerToDCM(euler, R);
  rtToPose(R, t, Pose);
227 228 229 230 231 232 233 234
}

/* Fast way to look up the duplicates
duplicates is pre-allocated
make sure that the max element in array will not exceed maxElement
*/
static hashtable_int* getHashtable(int* data, size_t length, int numMaxElement)
{
Bence Magyar's avatar
Bence Magyar committed
235 236 237 238 239 240 241 242
  hashtable_int* hashtable = hashtableCreate(static_cast<size_t>(numMaxElement*2), 0);
  for (size_t i = 0; i < length; i++)
  {
    const KeyType key = (KeyType)data[i];
    hashtableInsertHashed(hashtable, key+1, reinterpret_cast<void*>(i+1));
  }

  return hashtable;
243 244 245
}

// source point clouds are assumed to contain their normals
Hamdi Sahloul's avatar
Hamdi Sahloul committed
246
int ICP::registerModelToScene(const Mat& srcPC, const Mat& dstPC, double& residual, Matx44d& pose)
247
{
Bence Magyar's avatar
Bence Magyar committed
248 249
  int n = srcPC.rows;

250
  const bool useRobustReject = m_rejectionScale>0;
Bence Magyar's avatar
Bence Magyar committed
251 252 253

  Mat srcTemp = srcPC.clone();
  Mat dstTemp = dstPC.clone();
254
  Vec3d meanSrc, meanDst;
Bence Magyar's avatar
Bence Magyar committed
255 256
  computeMeanCols(srcTemp, meanSrc);
  computeMeanCols(dstTemp, meanDst);
257
  Vec3d meanAvg = 0.5 * (meanSrc + meanDst);
Bence Magyar's avatar
Bence Magyar committed
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
  subtractColumns(srcTemp, meanAvg);
  subtractColumns(dstTemp, meanAvg);

  double distSrc = computeDistToOrigin(srcTemp);
  double distDst = computeDistToOrigin(dstTemp);

  double scale = (double)n / ((distSrc + distDst)*0.5);

  srcTemp(cv::Range(0, srcTemp.rows), cv::Range(0,3)) *= scale;
  dstTemp(cv::Range(0, dstTemp.rows), cv::Range(0,3)) *= scale;

  Mat srcPC0 = srcTemp;
  Mat dstPC0 = dstTemp;

  // initialize pose
273
  pose = Matx44d::eye();
Bence Magyar's avatar
Bence Magyar committed
274 275 276 277 278 279 280 281 282 283 284 285 286

  Mat M = Mat::eye(4,4,CV_64F);

  double tempResidual = 0;


  // walk the pyramid
  for (int level = m_numLevels-1; level >=0; level--)
  {
    const double impact = 2;
    double div = pow((double)impact, (double)level);
    //double div2 = div*div;
    const int numSamples = cvRound((double)(n/(div)));
287 288
    const double TolP = m_tolerance*(double)(level+1)*(level+1);
    const int MaxIterationsPyr = cvRound((double)m_maxIterations/(level+1));
Bence Magyar's avatar
Bence Magyar committed
289 290

    // Obtain the sampled point clouds for this level: Also rotates the normals
291
    Mat srcPCT = transformPCPose(srcPC0, pose);
Bence Magyar's avatar
Bence Magyar committed
292 293 294

    const int sampleStep = cvRound((double)n/(double)numSamples);

295
    srcPCT = samplePCUniform(srcPCT, sampleStep);
Bence Magyar's avatar
Bence Magyar committed
296
    /*
297 298
    Tolga Birdal thinks that downsampling the scene points might decrease the accuracy.
    Hamdi Sahloul, however, noticed that accuracy increased (pose residual decreased slightly).
Bence Magyar's avatar
Bence Magyar committed
299
    */
300 301
    Mat dstPCS = samplePCUniform(dstPC0, sampleStep);
    void* flann = indexPCFlann(dstPCS);
Bence Magyar's avatar
Bence Magyar committed
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324

    double fval_old=9999999999;
    double fval_perc=0;
    double fval_min=9999999999;
    Mat Src_Moved = srcPCT.clone();

    int i=0;

    size_t numElSrc = (size_t)Src_Moved.rows;
    int sizesResult[2] = {(int)numElSrc, 1};
    float* distances = new float[numElSrc];
    int* indices = new int[numElSrc];

    Mat Indices(2, sizesResult, CV_32S, indices, 0);
    Mat Distances(2, sizesResult, CV_32F, distances, 0);

    // use robust weighting for outlier treatment
    int* indicesModel = new int[numElSrc];
    int* indicesScene = new int[numElSrc];

    int* newI = new int[numElSrc];
    int* newJ = new int[numElSrc];

325
    Matx44d PoseX = Matx44d::eye();
Bence Magyar's avatar
Bence Magyar committed
326 327

    while ( (!(fval_perc<(1+TolP) && fval_perc>(1-TolP))) && i<MaxIterationsPyr)
328
    {
329
      uint di=0, selInd = 0;
Bence Magyar's avatar
Bence Magyar committed
330 331 332 333 334

      queryPCFlann(flann, Src_Moved, Indices, Distances);

      for (di=0; di<numElSrc; di++)
      {
335
        newI[di] = di;
Bence Magyar's avatar
Bence Magyar committed
336 337 338
        newJ[di] = indices[di];
      }

339
      if (useRobustReject)
Bence Magyar's avatar
Bence Magyar committed
340 341 342 343 344 345 346
      {
        int numInliers = 0;
        float threshold = getRejectionThreshold(distances, Distances.rows, m_rejectionScale);
        Mat acceptInd = Distances<threshold;

        uchar *accPtr = (uchar*)acceptInd.data;
        for (int l=0; l<acceptInd.rows; l++)
347
        {
Bence Magyar's avatar
Bence Magyar committed
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
          if (accPtr[l])
          {
            newI[numInliers] = l;
            newJ[numInliers] = indices[l];
            numInliers++;
          }
        }
        numElSrc=numInliers;
      }

      // Step 2: Picky ICP
      // Among the resulting corresponding pairs, if more than one scene point p_i
      // is assigned to the same model point m_j, then select p_i that corresponds
      // to the minimum distance

363
      hashtable_int* duplicateTable = getHashtable(newJ, numElSrc, dstPCS.rows);
Bence Magyar's avatar
Bence Magyar committed
364 365 366 367 368 369 370 371

      for (di=0; di<duplicateTable->size; di++)
      {
        hashnode_i *node = duplicateTable->nodes[di];

        if (node)
        {
          // select the first node
372
          size_t idx = reinterpret_cast<size_t>(node->data)-1, dn=0;
Bence Magyar's avatar
Bence Magyar committed
373
          int dup = (int)node->key-1;
374
          size_t minIdxD = idx;
Bence Magyar's avatar
Bence Magyar committed
375 376 377 378
          float minDist = distances[idx];

          while ( node )
          {
379
            idx = reinterpret_cast<size_t>(node->data)-1;
Bence Magyar's avatar
Bence Magyar committed
380 381

            if (distances[idx] < minDist)
382
            {
Bence Magyar's avatar
Bence Magyar committed
383 384
              minDist = distances[idx];
              minIdxD = idx;
385
            }
Bence Magyar's avatar
Bence Magyar committed
386 387 388 389 390 391 392 393

            node = node->next;
            dn++;
          }

          indicesModel[ selInd ] = newI[ minIdxD ];
          indicesScene[ selInd ] = dup ;
          selInd++;
394
        }
Bence Magyar's avatar
Bence Magyar committed
395 396 397 398
      }

      hashtableDestroy(duplicateTable);

399
      if (selInd >= 6)
Bence Magyar's avatar
Bence Magyar committed
400 401
      {

402 403
        Mat Src_Match = Mat(selInd, srcPCT.cols, CV_64F);
        Mat Dst_Match = Mat(selInd, srcPCT.cols, CV_64F);
Bence Magyar's avatar
Bence Magyar committed
404 405 406 407 408

        for (di=0; di<selInd; di++)
        {
          const int indModel = indicesModel[di];
          const int indScene = indicesScene[di];
409
          const float *srcPt = srcPCT.ptr<float>(indModel);
410
          const float *dstPt = dstPCS.ptr<float>(indScene);
411 412
          double *srcMatchPt = Src_Match.ptr<double>(di);
          double *dstMatchPt = Dst_Match.ptr<double>(di);
Bence Magyar's avatar
Bence Magyar committed
413 414 415 416 417 418 419 420 421
          int ci=0;

          for (ci=0; ci<srcPCT.cols; ci++)
          {
            srcMatchPt[ci] = (double)srcPt[ci];
            dstMatchPt[ci] = (double)dstPt[ci];
          }
        }

422 423 424 425 426
        Vec3d rpy, t;
        minimizePointToPlaneMetric(Src_Match, Dst_Match, rpy, t);
        if (cvIsNaN(cv::trace(rpy)) || cvIsNaN(cv::norm(t)))
          break;
        getTransformMat(rpy, t, PoseX);
Bence Magyar's avatar
Bence Magyar committed
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444
        Src_Moved = transformPCPose(srcPCT, PoseX);

        double fval = cv::norm(Src_Match, Dst_Match)/(double)(Src_Moved.rows);

        // Calculate change in error between iterations
        fval_perc=fval/fval_old;

        // Store error value
        fval_old=fval;

        if (fval < fval_min)
          fval_min = fval;
      }
      else
        break;

      i++;

445
    }
Bence Magyar's avatar
Bence Magyar committed
446

447
    pose = PoseX * pose;
448
    residual = tempResidual;
Bence Magyar's avatar
Bence Magyar committed
449 450 451 452 453 454 455 456 457

    delete[] newI;
    delete[] newJ;
    delete[] indicesModel;
    delete[] indicesScene;
    delete[] distances;
    delete[] indices;

    tempResidual = fval_min;
458
    destroyFlann(flann);
Bence Magyar's avatar
Bence Magyar committed
459 460
  }

461 462 463 464 465
  Matx33d Rpose;
  Vec3d Cpose;
  poseToRT(pose, Rpose, Cpose);
  Cpose = Cpose / scale + meanAvg - Rpose * meanAvg;
  rtToPose(Rpose, Cpose, pose);
Bence Magyar's avatar
Bence Magyar committed
466

467
  residual = tempResidual;
Bence Magyar's avatar
Bence Magyar committed
468 469

  return 0;
470 471 472
}

// source point clouds are assumed to contain their normals
473
int ICP::registerModelToScene(const Mat& srcPC, const Mat& dstPC, std::vector<Pose3DPtr>& poses)
474
{
475 476 477 478
  #if defined _OPENMP
  #pragma omp parallel for
  #endif
  for (int i=0; i<(int)poses.size(); i++)
Bence Magyar's avatar
Bence Magyar committed
479
  {
Hamdi Sahloul's avatar
Hamdi Sahloul committed
480
    Matx44d poseICP = Matx44d::eye();
481 482
    Mat srcTemp = transformPCPose(srcPC, poses[i]->pose);
    registerModelToScene(srcTemp, dstPC, poses[i]->residual, poseICP);
483
    poses[i]->appendPose(poseICP);
Bence Magyar's avatar
Bence Magyar committed
484 485
  }
  return 0;
486 487 488 489 490
}

} // namespace ppf_match_3d

} // namespace cv