Mercurial > hg > nsaunier > traffic-intelligence

comparison c/feature-based-tracking.cpp @ 800:2cade72d75ad dev

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
modified so there is no tracking outside of the mask (does not continue if features can still be matched) as requested
author Nicolas Saunier <nicolas.saunier@polymtl.ca>
date Tue, 31 May 2016 17:06:41 -0400
parents 85af65b6d531
children d3e8dd9f3ca4 daa992ac6b44
comparison
equal deleted inserted replaced
799:0662c87a61c9 800:2cade72d75ad
60 BOOST_FOREACH(FeatureTrajectoryPtr f, features) f->write(db, positionsTableName, velocitiesTableName); 60 BOOST_FOREACH(FeatureTrajectoryPtr f, features) f->write(db, positionsTableName, velocitiesTableName);
61 features.clear(); 61 features.clear();
62 } 62 }
63 63
64 void trackFeatures(const KLTFeatureTrackingParameters& params) { 64 void trackFeatures(const KLTFeatureTrackingParameters& params) {
65 // BriefDescriptorExtractor brief(32);
66 // const int DESIRED_FTRS = 500;
67 // GridAdaptedFeatureDetector detector(new FastFeatureDetector(10, true), DESIRED_FTRS, 4, 4);
68
69 Mat homography = ::loadMat(params.homographyFilename, " "); 65 Mat homography = ::loadMat(params.homographyFilename, " ");
70 Mat invHomography; 66 Mat invHomography;
71 if (params.display && !homography.empty()) 67 if (params.display && !homography.empty())
72 invHomography = homography.inv(); 68 invHomography = homography.inv();
73 69
172 if (frame.size() != videoSize) { 168 if (frame.size() != videoSize) {
173 cout << "Different frame size " << frameNum << ", breaking ([" << frame.size().width << "x" << frame.size().height << "])" << endl; 169 cout << "Different frame size " << frameNum << ", breaking ([" << frame.size().width << "x" << frame.size().height << "])" << endl;
174 break; 170 break;
175 } 171 }
176 } 172 }
177 173
178
179 cvtColor(frame, currentFrameBW, CV_RGB2GRAY); 174 cvtColor(frame, currentFrameBW, CV_RGB2GRAY);
180 175
181 if (!prevPts.empty()) { 176 if (!prevPts.empty()) {
182 currPts.clear(); 177 currPts.clear();
183 calcOpticalFlowPyrLK(previousFrameBW, currentFrameBW, prevPts, currPts, status, errors, window, params.pyramidLevel, TermCriteria(static_cast<int>(TermCriteria::COUNT)+static_cast<int>(TermCriteria::EPS) /* = 3 */, params.maxNumberTrackingIterations, params.minTrackingError), /* int flags = */ 0, params.minFeatureEigThreshold); 178 calcOpticalFlowPyrLK(previousFrameBW, currentFrameBW, prevPts, currPts, status, errors, window, params.pyramidLevel, TermCriteria(static_cast<int>(TermCriteria::COUNT)+static_cast<int>(TermCriteria::EPS) /* = 3 */, params.maxNumberTrackingIterations, params.minTrackingError), /* int flags = */ 0, params.minFeatureEigThreshold);
186 std::vector<Point2f> trackedPts; 181 std::vector<Point2f> trackedPts;
187 std::vector<FeaturePointMatch>::iterator iter = featurePointMatches.begin(); 182 std::vector<FeaturePointMatch>::iterator iter = featurePointMatches.begin();
188 while (iter != featurePointMatches.end()) { 183 while (iter != featurePointMatches.end()) {
189 bool deleteFeature = false; 184 bool deleteFeature = false;
190 185
191 if (status[iter->pointNum]) { 186 if (status[iter->pointNum] && (mask.at<uchar>(static_cast<int>(round(currPts[iter->pointNum].y)), static_cast<int>(round(currPts[iter->pointNum].x))) != 0)) {
192 iter->feature->addPoint(frameNum, currPts[iter->pointNum], homography); 187 iter->feature->addPoint(frameNum, currPts[iter->pointNum], homography);
193 188
194 deleteFeature = iter->feature->isDisplacementSmall(params.nDisplacements, minTotalFeatureDisplacement) 189 deleteFeature = iter->feature->isDisplacementSmall(params.nDisplacements, minTotalFeatureDisplacement)
195 || !iter->feature->isMotionSmooth(params.accelerationBound, params.deviationBound); 190 || !iter->feature->isMotionSmooth(params.accelerationBound, params.deviationBound);
196 if (deleteFeature) 191 if (deleteFeature)
197 iter->feature->shorten(); 192 iter->feature->shorten();
198 } else 193 } else
217 saveFeatures(lostFeatures, *trajectoryDB, "positions", "velocities"); 212 saveFeatures(lostFeatures, *trajectoryDB, "positions", "velocities");
218 213
219 if (params.display) { 214 if (params.display) {
220 BOOST_FOREACH(FeaturePointMatch fp, featurePointMatches) 215 BOOST_FOREACH(FeaturePointMatch fp, featurePointMatches)
221 fp.feature->draw(frame, invHomography, Colors::red()); 216 fp.feature->draw(frame, invHomography, Colors::red());
222 // object detection
223 // vector<Rect> locations;
224 // hog.detectMultiScale(frame, locations, 0, Size(8,8), Size(32,32), 1.05, 2);
225 // BOOST_FOREACH(Rect r, locations)
226 // rectangle(frame, r.tl(), r.br(), cv::Scalar(0,255,0), 3);
227 } 217 }
228 } 218 }
229 219
230 // adding new features, using mask around existing feature positions 220 // adding new features, using mask around existing feature positions
231 Mat featureMask = mask.clone(); 221 Mat featureMask = mask.clone();
232 for (unsigned int n=0;n<currPts.size(); n++) 222 for (unsigned int n=0;n<currPts.size(); n++)
233 for (int j=MAX(0, currPts[n].x-params.minFeatureDistanceKLT); j<MIN(videoSize.width, currPts[n].x+params.minFeatureDistanceKLT+1); j++) 223 for (int j=MAX(0, currPts[n].x-params.minFeatureDistanceKLT); j<MIN(videoSize.width, currPts[n].x+params.minFeatureDistanceKLT+1); j++)
234 for (int i=MAX(0, currPts[n].y-params.minFeatureDistanceKLT); i<MIN(videoSize.height, currPts[n].y+params.minFeatureDistanceKLT+1); i++) 224 for (int i=MAX(0, currPts[n].y-params.minFeatureDistanceKLT); i<MIN(videoSize.height, currPts[n].y+params.minFeatureDistanceKLT+1); i++)
235 featureMask.at<uchar>(i,j)=0; 225 featureMask.at<uchar>(i,j)=0;
236 goodFeaturesToTrack(currentFrameBW, newPts, params.maxNFeatures, params.featureQuality, params.minFeatureDistanceKLT, featureMask, params.blockSize, params.useHarrisDetector, params.k); 226 goodFeaturesToTrack(currentFrameBW, newPts, params.maxNFeatures, params.featureQuality, params.minFeatureDistanceKLT, featureMask, params.blockSize, params.useHarrisDetector, params.k);
237 BOOST_FOREACH(Point2f p, newPts) { //for (unsigned int i=0; i<newPts.size(); i++) { 227 BOOST_FOREACH(Point2f p, newPts) {
238 FeatureTrajectoryPtr f = FeatureTrajectoryPtr(new FeatureTrajectory(frameNum, p, homography)); 228 FeatureTrajectoryPtr f = FeatureTrajectoryPtr(new FeatureTrajectory(frameNum, p, homography));
239 featurePointMatches.push_back(FeaturePointMatch(f, currPts.size())); 229 featurePointMatches.push_back(FeaturePointMatch(f, currPts.size()));
240 currPts.push_back(p); 230 currPts.push_back(p);
241 } 231 }
242 232