PARP Research Group

Universidad de Murcia

---

examples/GPU-KLT+FLOW/demoSparse/demoSparse.cpp Go to the documentation of this file. /* * Copyright (C) 2011, 2012. PARP Research Group. * <http://perception.inf.um.es> * University of Murcia, Spain. * * This file is part of the QVision library. * * QVision is free software: you can redistribute it and/or modify * it under the terms of the GNU Lesser General Public License as * published by the Free Software Foundation, version 3 of the License. * * QVision is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with QVision. If not, see <http://www.gnu.org/licenses/>. */ #include <GL/glew.h> #include<QVApplication> #include<QVDefaultGUI> #include<QVVideoReaderBlock> #include<QVImageCanvas> #include<QVProcessingBlock> #include<QVImage> #include<qvipp.h> #include <qvgpukltflow.h> #include <QVKLTFeatureTracking> #ifdef Q_WS_X11 #include <X11/Xlib.h> // For XInitThreads function (called in main()) #endif // Needed to treat new types as QVision properties: typedef QHash< int, QVKLTFeatureTracking> TrackingHash; Q_DECLARE_METATYPE(TrackingHash); //typedef QHash<int,QVKLTTrackerFeature> FeatureHash; //Q_DECLARE_METATYPE(FeatureHash); // Canvas to draw left and right images, and matchings: /*class MatchingCanvas: public QVImageCanvas { public: MatchingCanvas(const QString name=QString(),QWidget *parent=0) : QVImageCanvas(name,parent) { addProperty<QHash<int,QVKLTTrackerFeature> >("featuresInput", inputFlag); addProperty<QHash<int,QVKLTTrackerFeature> >("featuresKeyframe", inputFlag); }; QColor hashColor(float value) { return( QColor( (int) (123456*value)%255, (int) (23456*value)%255, (int) (3456*value)%255) ); } void custom_viewer() { QHash<int,QVKLTTrackerFeature> featuresInput = getPropertyValue<QHash<int,QVKLTTrackerFeature> >("featuresInput"), featuresKeyframe = getPropertyValue<QHash<int,QVKLTTrackerFeature> >("featuresKeyframe"); getQVPainter()->setBackgroundMode(Qt::OpaqueMode); int width = this->getSize().width()/2; getQVPainter()->setPen(Qt::red); getQVPainter()->drawText(25,25,QString("%1 matchings").arg(featuresInput.size())); QHash<int, QVKLTTrackerFeature>::iterator i; for(i = featuresInput.begin(); i != featuresInput.end(); i++) { QVKLTTrackerFeature ftInput = i.value(); QVKLTTrackerFeature ftKey = featuresKeyframe.value(i.key()); getQVPainter()->setPen(hashColor(ftKey.getPosition().x()/width)); getQVPainter()->drawLine(ftKey.getPosition(),ftInput.getPosition()+QPointF(width,0)); } }; };*/ class MyQVImageCanvas: public QVImageCanvas { public: MyQVImageCanvas(const QString name=QString(),QWidget *parent=0) : QVImageCanvas(name,parent) { addProperty<TrackingHash>("Trackings", inputFlag); }; // Draw tracked features: void custom_viewer() { QHash<int, QVKLTFeatureTracking> trackings = getPropertyValue<TrackingHash>("Trackings"); getQVPainter()->setBackgroundMode(Qt::OpaqueMode); foreach(QVKLTFeatureTracking ft, trackings) { QList<QVKLTTrackerFeature> tracking = ft.getFeatureList(); // Draw current position: if(tracking.size()==1) // New feature getQVPainter()->setPen(Qt::red); else getQVPainter()->setPen(Qt::green); // Tracked feature getQVPainter()->drawEllipse(tracking.last().getPosition(), 2.0, 2.0); // Draw trajectory: getQVPainter()->setPen(Qt::blue); for(int j=0; j < tracking.size()-1; j++) { getQVPainter()->drawLine(tracking[j].getPosition(), tracking[j+1].getPosition()); } } getQVPainter()->drawText(25,25,QString("%1 features").arg(trackings.size())); }; }; class MyProcessingBlock : public QVProcessingBlock { private: QVKLTTracker *tracker,*matchTracker; QVTrackingManager* trackingManager; QHash<int,QVKLTTrackerFeature> features; QVImage<uChar,1> imageIn,keyImage; public: MyProcessingBlock(QString name): QVProcessingBlock(name) { tracker = 0; matchTracker = 0; trackingManager = 0; addProperty<QVImage<uChar,1> >("imageIn",inputFlag|outputFlag); addProperty<int>("nTrackedFrames",inputFlag,10,"Number of frames before redetection",1,30); addProperty<bool>("trackWithGain",inputFlag,true,"Wether track using gain or not"); addProperty<int>("featuresWidth",inputFlag,64,"Features width",8,128); addProperty<int>("featuresHeight",inputFlag,64,"Features height",8,128); addProperty<int>("nIterations",inputFlag,10,"number of KLT iterations",10,30); addProperty<int>("nLevels",inputFlag,4,"Number of levels of pyramid",1,4); addProperty<int>("levelSkip",inputFlag,1,"Levels of pyramid to skip",1,3); addProperty<int>("windowWidth",inputFlag,7,"Window width used by KLT",1,21); addProperty<float>("trackBorderMargin",inputFlag,20.0,"Border margin for tracking features",10.0,50.0); addProperty<float>("detectBorderMargin",inputFlag,20.0,"Border margin for detecting features",1.0,100.0); addProperty<float>("convergenceThreshold",inputFlag,0.1,"Convergence threshold for KLT",0.01,0.5); addProperty<float>("SSD_Threshold",inputFlag,5000.0,"SSD threshold for KLT",1000.0,10000.0); addProperty<int>("minDistance",inputFlag,7,"Minimum distance between features",3,25); addProperty<float>("minCornerness",inputFlag,50.0,"Minimum value for cornerness",20.0,100.0); addProperty<QHash<int,QVKLTFeatureTracking> > ("Trackings",outputFlag); addProperty<QVImage<uChar,1> >("Matching images",outputFlag); addProperty< QList<QPointFMatching> >("Matching correspondences", outputFlag); // ------------------------ //addProperty<QHash<int,QVKLTTrackerFeature> >("featuresInput",outputFlag); //addProperty<QHash<int,QVKLTTrackerFeature> >("featuresKeyframe",outputFlag); addTrigger("Follow from key frame"); }; void processTrigger(QString name) { if(name == QString("Follow from key frame")) { keyImage = imageIn; } } // Function to reset tracker when any parameter changed: void ResetTracker(int width, int height, bool trackWithGain, int featuresWidth, int featuresHeight, int nIterations, int nLevels, int levelSkip,int windowWidth, float trackBorderMargin, float detectBorderMargin, float convergenceThreshold, float SSD_Threshold, int minDistance, float minCornerness) { // If it is not the first time, delete former tracker, and then create new one: if(tracker != 0) { delete tracker; } tracker = new QVKLTTracker(width, height, trackWithGain, featuresWidth, featuresHeight, nIterations, nLevels, levelSkip, windowWidth, trackBorderMargin, detectBorderMargin, convergenceThreshold, SSD_Threshold, minDistance, minCornerness); if(matchTracker != 0) { delete matchTracker; } matchTracker = new QVKLTTracker(width, height, trackWithGain, featuresWidth, featuresHeight, nIterations, nLevels, levelSkip, windowWidth, trackBorderMargin, detectBorderMargin, convergenceThreshold, SSD_Threshold, minDistance, minCornerness); // Same for tracking manager: if (trackingManager != 0){ delete trackingManager; } trackingManager = new QVTrackingManager(true); } void iterate() { bool reset = false; // Initialize only in first iteration (must be done here in the processing block thread): if(getIteration() == 0) { InitGLEWCg(); reset = true; } // Read input properties:va static bool trackWithGain; static int width = 0, height = 0, featuresWidth = 0, featuresHeight = 0, nIterations = 0, nLevels = 0, levelSkip = 0, windowWidth = 0, minDistance = 0; static float trackBorderMargin = 0.0, detectBorderMargin = 0.0, convergenceThreshold = 0.0, SSD_Threshold = 0.0, minCornerness = 0.0; int nTrackedFrames = getPropertyValue<float>("nTrackedFrames"); imageIn = getPropertyValue<QVImage<uChar,1> >("imageIn"); if ( (width != imageIn.getCols()) or (height != imageIn.getRows()) or (trackWithGain != getPropertyValue<bool>("trackWithGain")) or (featuresWidth != getPropertyValue<int>("featuresWidth")) or (featuresHeight != getPropertyValue<int>("featuresHeight")) or (nIterations != getPropertyValue<int>("nIterations")) or (nLevels != getPropertyValue<int>("nLevels")) or (levelSkip != getPropertyValue<int>("levelSkip")) or (windowWidth != getPropertyValue<int>("windowWidth")) or (trackBorderMargin != getPropertyValue<float>("trackBorderMargin")) or (detectBorderMargin != getPropertyValue<float>("detectBorderMargin")) or (convergenceThreshold != getPropertyValue<float>("convergenceThreshold")) or (SSD_Threshold != getPropertyValue<float>("SSD_Threshold")) or (minDistance != getPropertyValue<int>("minDistance")) or (minCornerness != getPropertyValue<float>("minCornerness")) ) { width = imageIn.getCols(); height = imageIn.getRows(); trackWithGain = getPropertyValue<bool>("trackWithGain"); featuresWidth = getPropertyValue<int>("featuresWidth"); featuresHeight = getPropertyValue<int>("featuresHeight"); nIterations = getPropertyValue<int>("nIterations"); nLevels = getPropertyValue<int>("nLevels"); levelSkip = getPropertyValue<int>("levelSkip"); windowWidth = getPropertyValue<int>("windowWidth"); trackBorderMargin = getPropertyValue<float>("trackBorderMargin"); detectBorderMargin = getPropertyValue<float>("detectBorderMargin"); convergenceThreshold = getPropertyValue<float>("convergenceThreshold"); SSD_Threshold = getPropertyValue<float>("SSD_Threshold"); minDistance = getPropertyValue<int>("minDistance"); minCornerness = getPropertyValue<float>("minCornerness"); reset = true; } /* if(width != imageIn.getCols()) { width = imageIn.getCols(); reset = true; } if(height != imageIn.getRows()) { height = imageIn.getRows(); reset = true; } if(trackWithGain != getPropertyValue<bool>("trackWithGain")) { trackWithGain = getPropertyValue<bool>("trackWithGain"); reset = true; } if(featuresWidth != getPropertyValue<int>("featuresWidth")) { featuresWidth = getPropertyValue<int>("featuresWidth"); reset = true; } if(featuresHeight != getPropertyValue<int>("featuresHeight")) { featuresHeight = getPropertyValue<int>("featuresHeight"); reset = true; } if(nIterations != getPropertyValue<int>("nIterations")) { nIterations = getPropertyValue<int>("nIterations"); reset = true; } if(nLevels != getPropertyValue<int>("nLevels")) { nLevels = getPropertyValue<int>("nLevels"); reset = true; } if(levelSkip != getPropertyValue<int>("levelSkip")) { levelSkip = getPropertyValue<int>("levelSkip"); reset = true; } if(windowWidth != getPropertyValue<int>("windowWidth")) { windowWidth = getPropertyValue<int>("windowWidth"); reset = true; } if(trackBorderMargin != getPropertyValue<float>("trackBorderMargin")) { trackBorderMargin = getPropertyValue<float>("trackBorderMargin"); reset = true; } if(detectBorderMargin != getPropertyValue<float>("detectBorderMargin")) { detectBorderMargin = getPropertyValue<float>("detectBorderMargin"); reset = true; } if(convergenceThreshold != getPropertyValue<float>("convergenceThreshold")) { convergenceThreshold = getPropertyValue<float>("convergenceThreshold"); reset = true; } if(SSD_Threshold != getPropertyValue<float>("SSD_Threshold")) { SSD_Threshold = getPropertyValue<float>("SSD_Threshold"); reset = true; } if(minDistance != getPropertyValue<int>("minDistance")) { minDistance = getPropertyValue<int>("minDistance"); reset = true; } if(minCornerness != getPropertyValue<float>("minCornerness")) { minCornerness = getPropertyValue<float>("minCornerness"); reset = true; }*/ timeFlag("Read properties"); // Reset tracker if needed (some configuration parameter of the tracker changed): if(reset) { ResetTracker(width, height, trackWithGain, featuresWidth, featuresHeight, nIterations, nLevels, levelSkip, windowWidth, trackBorderMargin, detectBorderMargin, convergenceThreshold, SSD_Threshold, minDistance, minCornerness); features.clear(); tracker->detect(imageIn, features); } timeFlag("Reset tracker"); // Otherwise, track features (redetecting new ones every nTrackedFrames): if(not reset) { if (getIteration() % nTrackedFrames == 0) //{ // Redetect every nTrackedFrames frames: tracker->redetect(imageIn, features); //} else //{ // Simply track set of features in rest of frames: tracker->track(imageIn, features); //} } timeFlag("GPU tracking"); // Update tracking manager with features: trackingManager->updateTrackings(&features); timeFlag("Tracking manager update"); // Publish results: setPropertyValue<TrackingHash>("Trackings", trackingManager->getTrackings()); timeFlag("Write properties"); if(keyImage.getSize() == imageIn.getSize()) { QVImage<uChar,1> imagesLeftAndRight; imagesLeftAndRight = QVImage<uChar,1>(imageIn.getCols()*2,imageIn.getRows()); Copy(keyImage,imagesLeftAndRight,QPoint(0, 0)); Copy(imageIn,imagesLeftAndRight,QPoint(imageIn.getCols(),0)); imagesLeftAndRight.resetROI(); QHash<int,QVKLTTrackerFeature> featuresKeyframe, featuresInput; matchTracker->detect(keyImage, featuresKeyframe); featuresInput = featuresKeyframe; matchTracker->track(imageIn, featuresInput); // ----------------------- QList<QPointFMatching> correspondences; const QPointF offset(keyImage.getCols(), 0); for(QHash<int, QVKLTTrackerFeature>::iterator i = featuresInput.begin(); i != featuresInput.end(); i++) { QVKLTTrackerFeature ftInput = i.value(); QVKLTTrackerFeature ftKey = featuresKeyframe.value(i.key()); correspondences << QPointFMatching(ftKey.getPosition(), ftInput.getPosition()+offset); } // ----------------------- setPropertyValue< QList<QPointFMatching> >("Matching correspondences", correspondences); setPropertyValue<QVImage<uChar,1> >("Matching images", imagesLeftAndRight); //setPropertyValue<QHash<int,QVKLTTrackerFeature> >("featuresKeyframe", featuresKeyframe); //setPropertyValue<QHash<int,QVKLTTrackerFeature> >("featuresInput", featuresInput); } } }; int main( int argc, char** argv) { initShadersEnvironment(); #ifdef Q_WS_X11 // Needed to make X11 calls thread-safe. // (Explanation in http://doc.qt.nokia.com/qq/qq06-glimpsing.html#writingmultithreadedglapplications) XInitThreads(); #endif QVApplication app(argc,argv); QVDefaultGUI gui; QVVideoReaderBlock camera("Camera"); MyProcessingBlock processingBlock("My processing block"); camera.linkProperty(&processingBlock,"imageIn"); MyQVImageCanvas imageCanvas("Canvas"); processingBlock.linkProperty("imageIn",&imageCanvas); processingBlock.linkProperty("Trackings", &imageCanvas, "Trackings"); QVImageCanvas referenceFrame("Matchings"); processingBlock.linkProperty("Matching images", referenceFrame); processingBlock.linkProperty("Matching correspondences", referenceFrame); referenceFrame.setRadius("Matching correspondences", 2); return app.exec(); }

---

QVision framework. PARP research group. Copyright © 2007, 2008, 2009, 2010, 2011.