Point Cloud Library (PCL)
1.7.0
|
00001 /* 00002 * Software License Agreement (BSD License) 00003 * 00004 * Point Cloud Library (PCL) - www.pointclouds.org 00005 * Copyright (c) 2009, Willow Garage, Inc. 00006 * Copyright (c) 2012-, Open Perception, Inc. 00007 * 00008 * All rights reserved. 00009 * 00010 * Redistribution and use in source and binary forms, with or without 00011 * modification, are permitted provided that the following conditions 00012 * are met: 00013 * 00014 * * Redistributions of source code must retain the above copyright 00015 * notice, this list of conditions and the following disclaimer. 00016 * * Redistributions in binary form must reproduce the above 00017 * copyright notice, this list of conditions and the following 00018 * disclaimer in the documentation and/or other materials provided 00019 * with the distribution. 00020 * * Neither the name of the copyright holder(s) nor the names of its 00021 * contributors may be used to endorse or promote products derived 00022 * from this software without specific prior written permission. 00023 * 00024 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 00025 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 00026 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 00027 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 00028 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 00029 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 00030 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 00031 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 00032 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 00033 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 00034 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 00035 * POSSIBILITY OF SUCH DAMAGE. 00036 * 00037 * $Id$ 00038 * 00039 */ 00040 00041 #ifndef PCL_SAMPLE_CONSENSUS_IMPL_MLESAC_H_ 00042 #define PCL_SAMPLE_CONSENSUS_IMPL_MLESAC_H_ 00043 00044 #include <pcl/sample_consensus/mlesac.h> 00045 #include <pcl/point_types.h> 00046 00047 ////////////////////////////////////////////////////////////////////////// 00048 template <typename PointT> bool 00049 pcl::MaximumLikelihoodSampleConsensus<PointT>::computeModel (int debug_verbosity_level) 00050 { 00051 // Warn and exit if no threshold was set 00052 if (threshold_ == std::numeric_limits<double>::max()) 00053 { 00054 PCL_ERROR ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] No threshold set!\n"); 00055 return (false); 00056 } 00057 00058 iterations_ = 0; 00059 double d_best_penalty = std::numeric_limits<double>::max(); 00060 double k = 1.0; 00061 00062 std::vector<int> best_model; 00063 std::vector<int> selection; 00064 Eigen::VectorXf model_coefficients; 00065 std::vector<double> distances; 00066 00067 // Compute sigma - remember to set threshold_ correctly ! 00068 sigma_ = computeMedianAbsoluteDeviation (sac_model_->getInputCloud (), sac_model_->getIndices (), threshold_); 00069 if (debug_verbosity_level > 1) 00070 PCL_DEBUG ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] Estimated sigma value: %f.\n", sigma_); 00071 00072 // Compute the bounding box diagonal: V = sqrt (sum (max(pointCloud) - min(pointCloud)^2)) 00073 Eigen::Vector4f min_pt, max_pt; 00074 getMinMax (sac_model_->getInputCloud (), sac_model_->getIndices (), min_pt, max_pt); 00075 max_pt -= min_pt; 00076 double v = sqrt (max_pt.dot (max_pt)); 00077 00078 int n_inliers_count = 0; 00079 size_t indices_size; 00080 unsigned skipped_count = 0; 00081 // supress infinite loops by just allowing 10 x maximum allowed iterations for invalid model parameters! 00082 const unsigned max_skip = max_iterations_ * 10; 00083 00084 // Iterate 00085 while (iterations_ < k && skipped_count < max_skip) 00086 { 00087 // Get X samples which satisfy the model criteria 00088 sac_model_->getSamples (iterations_, selection); 00089 00090 if (selection.empty ()) break; 00091 00092 // Search for inliers in the point cloud for the current plane model M 00093 if (!sac_model_->computeModelCoefficients (selection, model_coefficients)) 00094 { 00095 //iterations_++; 00096 ++ skipped_count; 00097 continue; 00098 } 00099 00100 // Iterate through the 3d points and calculate the distances from them to the model 00101 sac_model_->getDistancesToModel (model_coefficients, distances); 00102 00103 if (distances.empty ()) 00104 { 00105 //iterations_++; 00106 ++skipped_count; 00107 continue; 00108 } 00109 00110 // Use Expectiation-Maximization to find out the right value for d_cur_penalty 00111 // ---[ Initial estimate for the gamma mixing parameter = 1/2 00112 double gamma = 0.5; 00113 double p_outlier_prob = 0; 00114 00115 indices_size = sac_model_->getIndices ()->size (); 00116 std::vector<double> p_inlier_prob (indices_size); 00117 for (int j = 0; j < iterations_EM_; ++j) 00118 { 00119 // Likelihood of a datum given that it is an inlier 00120 for (size_t i = 0; i < indices_size; ++i) 00121 p_inlier_prob[i] = gamma * exp (- (distances[i] * distances[i] ) / 2 * (sigma_ * sigma_) ) / 00122 (sqrt (2 * M_PI) * sigma_); 00123 00124 // Likelihood of a datum given that it is an outlier 00125 p_outlier_prob = (1 - gamma) / v; 00126 00127 gamma = 0; 00128 for (size_t i = 0; i < indices_size; ++i) 00129 gamma += p_inlier_prob [i] / (p_inlier_prob[i] + p_outlier_prob); 00130 gamma /= static_cast<double>(sac_model_->getIndices ()->size ()); 00131 } 00132 00133 // Find the log likelihood of the model -L = -sum [log (pInlierProb + pOutlierProb)] 00134 double d_cur_penalty = 0; 00135 for (size_t i = 0; i < indices_size; ++i) 00136 d_cur_penalty += log (p_inlier_prob[i] + p_outlier_prob); 00137 d_cur_penalty = - d_cur_penalty; 00138 00139 // Better match ? 00140 if (d_cur_penalty < d_best_penalty) 00141 { 00142 d_best_penalty = d_cur_penalty; 00143 00144 // Save the current model/coefficients selection as being the best so far 00145 model_ = selection; 00146 model_coefficients_ = model_coefficients; 00147 00148 n_inliers_count = 0; 00149 // Need to compute the number of inliers for this model to adapt k 00150 for (size_t i = 0; i < distances.size (); ++i) 00151 if (distances[i] <= 2 * sigma_) 00152 n_inliers_count++; 00153 00154 // Compute the k parameter (k=log(z)/log(1-w^n)) 00155 double w = static_cast<double> (n_inliers_count) / static_cast<double> (sac_model_->getIndices ()->size ()); 00156 double p_no_outliers = 1 - pow (w, static_cast<double> (selection.size ())); 00157 p_no_outliers = (std::max) (std::numeric_limits<double>::epsilon (), p_no_outliers); // Avoid division by -Inf 00158 p_no_outliers = (std::min) (1 - std::numeric_limits<double>::epsilon (), p_no_outliers); // Avoid division by 0. 00159 k = log (1 - probability_) / log (p_no_outliers); 00160 } 00161 00162 ++iterations_; 00163 if (debug_verbosity_level > 1) 00164 PCL_DEBUG ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] Trial %d out of %d. Best penalty is %f.\n", iterations_, static_cast<int> (ceil (k)), d_best_penalty); 00165 if (iterations_ > max_iterations_) 00166 { 00167 if (debug_verbosity_level > 0) 00168 PCL_DEBUG ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] MLESAC reached the maximum number of trials.\n"); 00169 break; 00170 } 00171 } 00172 00173 if (model_.empty ()) 00174 { 00175 if (debug_verbosity_level > 0) 00176 PCL_DEBUG ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] Unable to find a solution!\n"); 00177 return (false); 00178 } 00179 00180 // Iterate through the 3d points and calculate the distances from them to the model again 00181 sac_model_->getDistancesToModel (model_coefficients_, distances); 00182 std::vector<int> &indices = *sac_model_->getIndices (); 00183 if (distances.size () != indices.size ()) 00184 { 00185 PCL_ERROR ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] Estimated distances (%zu) differs than the normal of indices (%zu).\n", distances.size (), indices.size ()); 00186 return (false); 00187 } 00188 00189 inliers_.resize (distances.size ()); 00190 // Get the inliers for the best model found 00191 n_inliers_count = 0; 00192 for (size_t i = 0; i < distances.size (); ++i) 00193 if (distances[i] <= 2 * sigma_) 00194 inliers_[n_inliers_count++] = indices[i]; 00195 00196 // Resize the inliers vector 00197 inliers_.resize (n_inliers_count); 00198 00199 if (debug_verbosity_level > 0) 00200 PCL_DEBUG ("[pcl::MaximumLikelihoodSampleConsensus::computeModel] Model: %zu size, %d inliers.\n", model_.size (), n_inliers_count); 00201 00202 return (true); 00203 } 00204 00205 ////////////////////////////////////////////////////////////////////////// 00206 template <typename PointT> double 00207 pcl::MaximumLikelihoodSampleConsensus<PointT>::computeMedianAbsoluteDeviation ( 00208 const PointCloudConstPtr &cloud, 00209 const boost::shared_ptr <std::vector<int> > &indices, 00210 double sigma) 00211 { 00212 std::vector<double> distances (indices->size ()); 00213 00214 Eigen::Vector4f median; 00215 // median (dist (x - median (x))) 00216 computeMedian (cloud, indices, median); 00217 00218 for (size_t i = 0; i < indices->size (); ++i) 00219 { 00220 pcl::Vector4fMapConst pt = cloud->points[(*indices)[i]].getVector4fMap (); 00221 Eigen::Vector4f ptdiff = pt - median; 00222 ptdiff[3] = 0; 00223 distances[i] = ptdiff.dot (ptdiff); 00224 } 00225 00226 std::sort (distances.begin (), distances.end ()); 00227 00228 double result; 00229 size_t mid = indices->size () / 2; 00230 // Do we have a "middle" point or should we "estimate" one ? 00231 if (indices->size () % 2 == 0) 00232 result = (sqrt (distances[mid-1]) + sqrt (distances[mid])) / 2; 00233 else 00234 result = sqrt (distances[mid]); 00235 return (sigma * result); 00236 } 00237 00238 ////////////////////////////////////////////////////////////////////////// 00239 template <typename PointT> void 00240 pcl::MaximumLikelihoodSampleConsensus<PointT>::getMinMax ( 00241 const PointCloudConstPtr &cloud, 00242 const boost::shared_ptr <std::vector<int> > &indices, 00243 Eigen::Vector4f &min_p, 00244 Eigen::Vector4f &max_p) 00245 { 00246 min_p.setConstant (FLT_MAX); 00247 max_p.setConstant (-FLT_MAX); 00248 min_p[3] = max_p[3] = 0; 00249 00250 for (size_t i = 0; i < indices->size (); ++i) 00251 { 00252 if (cloud->points[(*indices)[i]].x < min_p[0]) min_p[0] = cloud->points[(*indices)[i]].x; 00253 if (cloud->points[(*indices)[i]].y < min_p[1]) min_p[1] = cloud->points[(*indices)[i]].y; 00254 if (cloud->points[(*indices)[i]].z < min_p[2]) min_p[2] = cloud->points[(*indices)[i]].z; 00255 00256 if (cloud->points[(*indices)[i]].x > max_p[0]) max_p[0] = cloud->points[(*indices)[i]].x; 00257 if (cloud->points[(*indices)[i]].y > max_p[1]) max_p[1] = cloud->points[(*indices)[i]].y; 00258 if (cloud->points[(*indices)[i]].z > max_p[2]) max_p[2] = cloud->points[(*indices)[i]].z; 00259 } 00260 } 00261 00262 ////////////////////////////////////////////////////////////////////////// 00263 template <typename PointT> void 00264 pcl::MaximumLikelihoodSampleConsensus<PointT>::computeMedian ( 00265 const PointCloudConstPtr &cloud, 00266 const boost::shared_ptr <std::vector<int> > &indices, 00267 Eigen::Vector4f &median) 00268 { 00269 // Copy the values to vectors for faster sorting 00270 std::vector<float> x (indices->size ()); 00271 std::vector<float> y (indices->size ()); 00272 std::vector<float> z (indices->size ()); 00273 for (size_t i = 0; i < indices->size (); ++i) 00274 { 00275 x[i] = cloud->points[(*indices)[i]].x; 00276 y[i] = cloud->points[(*indices)[i]].y; 00277 z[i] = cloud->points[(*indices)[i]].z; 00278 } 00279 std::sort (x.begin (), x.end ()); 00280 std::sort (y.begin (), y.end ()); 00281 std::sort (z.begin (), z.end ()); 00282 00283 size_t mid = indices->size () / 2; 00284 if (indices->size () % 2 == 0) 00285 { 00286 median[0] = (x[mid-1] + x[mid]) / 2; 00287 median[1] = (y[mid-1] + y[mid]) / 2; 00288 median[2] = (z[mid-1] + z[mid]) / 2; 00289 } 00290 else 00291 { 00292 median[0] = x[mid]; 00293 median[1] = y[mid]; 00294 median[2] = z[mid]; 00295 } 00296 median[3] = 0; 00297 } 00298 00299 #define PCL_INSTANTIATE_MaximumLikelihoodSampleConsensus(T) template class PCL_EXPORTS pcl::MaximumLikelihoodSampleConsensus<T>; 00300 00301 #endif // PCL_SAMPLE_CONSENSUS_IMPL_MLESAC_H_ 00302