// get next hand position using Kalman Filter vector<Point2f> visionUtils::getPredictedHandPosition(Point2f currentPoint, int mode) { vector<Point2f> predictedPoint; // First predict, to update the internal statePre variable Mat prediction = KF.predict(); Point predictPt(prediction.at<float>(0),prediction.at<float>(1)); measurement(0) = currentPoint.x; measurement(1) = currentPoint.y; // The update phase Mat estimated = KF.correct(measurement); Point statePt(estimated.at<float>(0),estimated.at<float>(1)); Point measPt(measurement(0),measurement(1)); middlePointV = measPt; kalmanMiddlePointV = statePt; predictedPoint.push_back(middlePointV); predictedPoint.push_back(kalmanMiddlePointV); return predictedPoint; }
vector<Point> Widget::findKalmanCenters(vector<Point> dataInput){ vector<Point> kalmanCenters; if(dataInput.empty()){ errorMsg("Nao ha objetos a serem detectados!"); return kalmanCenters; } kalmanCenters.resize(dataInput.size()); KalmanFilter *KF; for(unsigned int i = 0; i < targets.size(); i++){ KF = targets[i]->KF; //apenas conversao de estruturas usadas Mat_<float> measurement(2,1); measurement(0) = dataInput[i].x; measurement(1) = dataInput[i].y; Mat estimated; Mat prediction = KF->predict(); if(detectionSucess[i] == true) { //caso a detecção tenha sido um sucesso jogamos a nova medida no filtro //ao chamar correct já estamos adicionando a nova medida ao filtro estimated = KF->correct(measurement); } else{ /*caso a medição tenha falhada realimentamos o filtro com a própria predição anterior*/ Mat_<float> predictedMeasurement(2,1); predictedMeasurement(0) = prediction.at<float>(0); predictedMeasurement(1) = prediction.at<float>(1); estimated = KF->correct(predictedMeasurement); KF->errorCovPre.copyTo(KF->errorCovPost); //copiar a covPre para covPos [dica de um usuário de algum fórum] } Point statePt(estimated.at<float>(0),estimated.at<float>(1)); kalmanCenters[i] = statePt; /*existe o centro medido pela previsão e o centro que o filtro de kalman acredita ser o real. O centro de kalman é uma ponderação das medidas e do modelo com conhecimento prévio de um erro aleatório*/ } return kalmanCenters; }
Point TrackShirt::kalmanTracker(Point centroid) { Mat prediction = KF.predict(); Point predictPt(prediction.at<float>(0),prediction.at<float>(1)); Mat_<float> measurement(2,1); measurement(0) = centroid.x; measurement(1) = centroid.y; //Point measPt(measurement(0),measurement(1)); Mat estimated = KF.correct(measurement); Point statePt(estimated.at<float>(0),estimated.at<float>(1)); //cout<<"kalman est"<<estimated<<endl; //cout<<"kalman pt"<<statePt.x<<statePt.y<<endl; return statePt; }
int main(int argc, char *argv[]) { //test pour savoir si l'utilisateur a renseigne un parametre if(argc <= 3) { std::cout<<"---------------------------------------"<<std::endl<< "Veuillez rentrer la methode choisie : "<<std::endl<< "- template_tracking" <<std::endl<< "- LK_tracking" <<std::endl<< "- farneback" <<std::endl<< "- camshift_kalman" <<std::endl<< "- background_lk" <<std::endl<< "- background_kalman" <<std::endl<<std::endl<< "Le type de format : " <<std::endl<< "- video" <<std::endl<< "- image" <<std::endl<<std::endl<< "Le nom du fichier d'input" <<std::endl<< "---------------------------------------"<<std::endl; std::exit(EXIT_FAILURE); } //------------------VARIABLES----------------------------------------------// //Variables communes choiceAlgo algo; formatVideo format; std::string inputFileName(argv[3]); int nbTrames = 501; double fps = 0; std::vector<cv::Mat> sequence; cv::Mat previousSequence; int nbPedestrians = 0; cv::HOGDescriptor hog; hog.setSVMDetector(cv::HOGDescriptor::getDefaultPeopleDetector()); std::vector<cv::Rect> detectedPedestrian; // HOG + Good feature to track + LK std::vector<std::vector<cv::Point2f>> featuresDetected; std::vector<std::vector<cv::Point2f>> previousFeaturesDetected; // HOG + Template tracking std::vector<cv::Rect> boxes; std::vector<cv::Rect> previousBoxes; // Optical flow farneback cv::Mat flow; cv::Mat imGray; cv::Mat imGrayPrev; //camshift and kalman filter std::vector<cv::MatND> backProj; std::vector<cv::Rect> roiHogDetected; std::vector<cv::Rect> roiCamShift; std::vector<bool> detected; std::vector<cv::Mat> hist; std::vector<cv::RotatedRect> rectCamShift; cv::Point2f rect_points[4]; //--------------------------------------------------------------------------------------// //Background substraction, pour le tracking LK et goodfeaturestotrack regarder au dessus trackingOption tracking; cv::Ptr<cv::BackgroundSubtractor> pMOG2; std::vector<cv::Rect> detectedPedestrianFiltered; cv::KalmanFilter KF(4,2,0,CV_32F); cv::Mat_<float> measurement(2,1); cv::Mat prediction; cv::Mat estimated; pMOG2 = cv::createBackgroundSubtractorMOG2(); //Avec ajout du Haar cascade classifier std::vector<std::vector<cv::Rect>> rect_upper_body; cv::CascadeClassifier classifier; std::string upper_body_cascade_name = "haarcascade_fullbody.xml"; if(!classifier.load(upper_body_cascade_name)) { std::cout<<"le fichier "<<upper_body_cascade_name<<" ne peut etre charge"<<std::endl; return -1; } //--------------------------------------------------------------------------------------// //Background substraction and Kalman bool initKalman = true; cv::KalmanFilter Kalman(4,2,0,CV_32F); cv::Mat_<float> measurmt(2,1); cv::Mat predict; cv::Mat estim; Kalman.transitionMatrix.at<float>(0,0) = 1; Kalman.transitionMatrix.at<float>(0,1) = 0; Kalman.transitionMatrix.at<float>(0,2) = 1; Kalman.transitionMatrix.at<float>(0,3) = 0; Kalman.transitionMatrix.at<float>(1,0) = 0; Kalman.transitionMatrix.at<float>(1,1) = 1; Kalman.transitionMatrix.at<float>(1,2) = 0; Kalman.transitionMatrix.at<float>(1,3) = 1; Kalman.transitionMatrix.at<float>(2,0) = 0; Kalman.transitionMatrix.at<float>(2,1) = 0; Kalman.transitionMatrix.at<float>(2,2) = 1; Kalman.transitionMatrix.at<float>(2,3) = 0; Kalman.transitionMatrix.at<float>(3,0) = 0; Kalman.transitionMatrix.at<float>(3,1) = 0; Kalman.transitionMatrix.at<float>(3,2) = 0; Kalman.transitionMatrix.at<float>(3,3) = 1; measurmt.setTo(cv::Scalar(0)); cv::setIdentity(Kalman.measurementMatrix); cv::setIdentity(Kalman.processNoiseCov, cv::Scalar::all(1e-4)); cv::setIdentity(Kalman.measurementNoiseCov, cv::Scalar::all(1e-1)); cv::setIdentity(Kalman.errorCovPost, cv::Scalar::all(.1)); cv::Rect rectK; cv::Rect prevRectK; //acquisition de la video algo = detectAlgo(std::string(argv[1])); format = detectFormat(std::string(argv[2])); //------------------VIDEO--------------------------------------------------// if(format == SEQUENCE_IMAGE) sequence.resize(nbTrames); else if(format == VIDEO) std::cout<<"video"<<std::endl; extractVideoData(sequence, format, inputFileName, nbTrames, fps); cv::namedWindow("Video", cv::WINDOW_AUTOSIZE); //------------------TRAITEMENT-VIDEO---------------------------------------// for(int i=0;i<nbTrames;i++) { if(i>0) previousSequence = sequence[i-1]; else previousSequence = sequence[i]; ///------------------HOG + Good Features to track + LK-----------------// if(algo == HOG_GOODFEATURESTOTRACK_LK) { if(i%20 == 0) { detectedPedestrian = hogDetection(sequence[i], hog); nbPedestrians = detectedPedestrian.size(); if(nbPedestrians != 0) { featuresDetected = featuresDetection(sequence[i], detectedPedestrian); previousFeaturesDetected.resize(featuresDetected.size()); previousFeaturesDetected = featuresDetected; } } else if(previousFeaturesDetected.size() != 0) { featuresDetected = lucasKanadeTracking(previousSequence, sequence[i], previousFeaturesDetected); previousFeaturesDetected.clear(); previousFeaturesDetected.resize(featuresDetected.size()); previousFeaturesDetected = featuresDetected; } //--------Representation-------------------- /* cv::Scalar myColor; for(size_t j=0;j<featuresDetected.size();j++) { if(j%3 == 0) myColor = cv::Scalar(0,0,cv::RNG().uniform(200,255)); else if(j%2 == 0) myColor = cv::Scalar(0,cv::RNG().uniform(200,255),0); else myColor = cv::Scalar(cv::RNG().uniform(200,255),0,0); for(size_t k=0;k<featuresDetected[j].size();k++) { cv::circle(sequence[i], featuresDetected[j][k], 1, myColor,-1); } } */ for(size_t j=0;j<featuresDetected.size();j++) { cv::rectangle(sequence[i], cv::boundingRect(featuresDetected[j]), cv::Scalar( 0, 0, 255), 2, 8, 0 ); } //affichage de la video cv::imshow("Video", sequence[i]); } ///------------------HOG + Template Tracking---------------------------// else if(algo == HOG_TEMPLATE_TRACKING) { if(i%20 == 0) { detectedPedestrian = hogDetection(sequence[i], hog); nbPedestrians = detectedPedestrian.size(); if(nbPedestrians != 0) { boxes = templateTracking(sequence[i], detectedPedestrian, CV_TM_CCORR_NORMED); previousBoxes.resize(boxes.size()); previousBoxes = boxes; } } else if(previousBoxes.size() != 0) { boxes = templateTracking(sequence[i], previousBoxes, CV_TM_CCORR_NORMED); previousBoxes.clear(); previousBoxes.resize(boxes.size()); previousBoxes = boxes; } //--------Representation-------------------- for(size_t j=0;j<boxes.size();j++) { cv::rectangle(sequence[i], boxes[j], cv::Scalar( 0, 0, 255), 2, 8, 0 ); } //affichage de la video cv::imshow("Video", sequence[i]); } ///------------------HOG + Optical Flow Farneback----------------------// else if(algo == OPT_FLOW_FARNEBACK) { if(i!=0) { flow = cv::Mat::zeros(sequence[i].size(), CV_32FC2); cv::cvtColor(sequence[i], imGray, CV_BGR2GRAY); cv::cvtColor(sequence[i-1], imGrayPrev, CV_BGR2GRAY); cv::calcOpticalFlowFarneback(imGrayPrev, imGray, flow, 0.5, 3, 15, 3, 5, 1.2, 0); //-----------------Representation------------------------------// drawOptFlowMap(flow, imGrayPrev, 16, CV_RGB(0, 255, 0)); //dessin test //affichage de la video cv::imshow("Video", imGrayPrev); } } ///--------------HOG+Camshift + Kalman Filter--------------------------// else if(algo == CAMSHIFT_KALMAN_FILTER) { //camshift if(i%20 == 0 && roiCamShift.size() == 0) { roiHogDetected = hogDetection(sequence[i], hog); refineROI(roiCamShift, detected, roiHogDetected); //test if(roiCamShift.size() != 0) { /* roiCamShift[0].x += 30; roiCamShift[0].width -= 60; roiCamShift[0].y += 40; roiCamShift[0].height -= 100; */ roiCamShift[0].x += roiCamShift[0].width/2; roiCamShift[0].width = roiCamShift[0].width/3; //roiCamShift[0].y += roiCamShift[0].height/2; roiCamShift[0].height = roiCamShift[0].height/3; } // } backProj = computeProbImage(sequence[i], roiCamShift, hist, detected); if (roiCamShift.size() != 0) cv::imshow("temp", backProj[0]); ///-------Test-Camshift--------------------/// rectCamShift.resize(roiCamShift.size()); for(size_t j=0;j<roiCamShift.size();j++) { /* std::cout<<roiCamShift[j]<<std::endl; cv::rectangle(backProj[j], roiCamShift[j], cv::Scalar( 255, 0, 0), 2, 8, 0 ); //DEBUG cv::imshow("before camshift", backProj[j]); cv::waitKey(0); */ cv::Rect rectMeanShift; rectMeanShift = roiCamShift[j]; cv::meanShift(backProj[j], rectMeanShift, cv::TermCriteria(cv::TermCriteria::EPS | cv::TermCriteria::COUNT, 10, 1)); cv::rectangle(sequence[i], rectMeanShift, cv::Scalar( 0, 255, 0), 2, 8, 0 ); rectCamShift[j] = cv::CamShift(backProj[j], roiCamShift[j], cv::TermCriteria(cv::TermCriteria::EPS | cv::TermCriteria::COUNT, 10, 1)); rectCamShift[j].points(rect_points); for(int k = 0; k < 4; k++) cv::line(sequence[i], rect_points[k], rect_points[(k+1)%4], cv::Scalar( 0, 0, 255), 2, 8); } ///----------------------------------------/// //-----------------Representation----------------------------------// //dessin du rectangle for(size_t j=0;j<roiCamShift.size();j++) cv::rectangle(sequence[i], roiCamShift[j], cv::Scalar( 255, 0, 0), 2, 8, 0 ); //affichage de la video cv::imshow("Video", sequence[i]); } ///------------------BACKGROUND-SUBSTRACTION---------------------------// else if(algo == BACKGROUND_LK) { if(i%10 == 0) //égal 0 pour le test { backgroundSubstractionDetection(sequence[i], detectedPedestrianFiltered, pMOG2, tracking); } if(tracking == GOOD_FEATURES_TO_TRACK) { featuresDetected.resize(detectedPedestrianFiltered.size()); featuresDetected = featuresDetection(sequence[i], detectedPedestrianFiltered); previousFeaturesDetected.resize(featuresDetected.size()); previousFeaturesDetected = featuresDetected; tracking = LUCAS_KANADE; KF.transitionMatrix.at<float>(0,0) = 1; KF.transitionMatrix.at<float>(0,1) = 0; KF.transitionMatrix.at<float>(0,2) = 1; KF.transitionMatrix.at<float>(0,3) = 0; KF.transitionMatrix.at<float>(1,0) = 0; KF.transitionMatrix.at<float>(1,1) = 1; KF.transitionMatrix.at<float>(1,2) = 0; KF.transitionMatrix.at<float>(1,3) = 1; KF.transitionMatrix.at<float>(2,0) = 0; KF.transitionMatrix.at<float>(2,1) = 0; KF.transitionMatrix.at<float>(2,2) = 1; KF.transitionMatrix.at<float>(2,3) = 0; KF.transitionMatrix.at<float>(3,0) = 0; KF.transitionMatrix.at<float>(3,1) = 0; KF.transitionMatrix.at<float>(3,2) = 0; KF.transitionMatrix.at<float>(3,3) = 1; measurement.setTo(cv::Scalar(0)); for(size_t j=0;j<featuresDetected.size();j++) { detectedPedestrianFiltered[j] = cv::boundingRect(featuresDetected[j]); } KF.statePre.at<float>(0) = rectCenter(detectedPedestrianFiltered[0]).x; KF.statePre.at<float>(1) = rectCenter(detectedPedestrianFiltered[0]).y; KF.statePre.at<float>(2) = 0; KF.statePre.at<float>(3) = 0; cv::setIdentity(KF.measurementMatrix); cv::setIdentity(KF.processNoiseCov, cv::Scalar::all(1e-4)); cv::setIdentity(KF.measurementNoiseCov, cv::Scalar::all(1e-1)); cv::setIdentity(KF.errorCovPost, cv::Scalar::all(.1)); } else if(tracking == LUCAS_KANADE) { for(size_t j=0;j<featuresDetected.size();j++) { detectedPedestrianFiltered[j] = cv::boundingRect(featuresDetected[j]); } featuresDetected = lucasKanadeTracking(previousSequence, sequence[i], previousFeaturesDetected); previousFeaturesDetected.clear(); previousFeaturesDetected.resize(featuresDetected.size()); previousFeaturesDetected = featuresDetected; prediction = KF.predict(); cv::Point predictPt(prediction.at<float>(0),prediction.at<float>(1)); // Get mouse point measurement(0) = rectCenter(detectedPedestrianFiltered[0]).x; measurement(1) = rectCenter(detectedPedestrianFiltered[0]).y; cv::Point measPt(measurement(0),measurement(1)); // The "correct" phase that is going to use the predicted value and our measurement cv::Mat estimated = KF.correct(measurement); cv::Point statePt(estimated.at<float>(0),estimated.at<float>(1)); //cv::circle(sequence[i], measPt, 1, cv::Scalar(0,255,0), 7, 24); //cv::circle(sequence[i], predictPt, 1, cv::Scalar(0,255,255), 7, 24); } //--------Representation-------------------- if(tracking != NOTHING_TO_TRACK) findUpperBody(classifier, sequence[i], detectedPedestrianFiltered, rect_upper_body); for(size_t j=0;j<featuresDetected.size();j++) { for(size_t k=0;k<rect_upper_body[j].size();k++) { cv::rectangle(sequence[i], rect_upper_body[j][k], cv::Scalar( 0, 255, 0), 2, 8, 0 ); } //detectedPedestrianFiltered[j] = cv::boundingRect(featuresDetected[j]); cv::rectangle(sequence[i], cv::boundingRect(featuresDetected[j]), cv::Scalar( 0, 0, 255), 2, 8, 0 ); } //affichage de la video cv::imshow("Video", sequence[i]); } else if(algo == BACKGROUND_KALMAN) { int refresh = 6; if(i%refresh == 0) { backgroundSubstractionDetection(sequence[i], detectedPedestrianFiltered, pMOG2, tracking); } if(initKalman && detectedPedestrianFiltered.size() != 0) { Kalman.statePre.at<float>(0) = rectCenter(detectedPedestrianFiltered[0]).x; Kalman.statePre.at<float>(1) = rectCenter(detectedPedestrianFiltered[0]).y; Kalman.statePre.at<float>(2) = 0; Kalman.statePre.at<float>(3) = 0; initKalman = false; } if(detectedPedestrianFiltered.size() != 0) { predict = Kalman.predict(); cv::Point predictPt(predict.at<float>(0),predict.at<float>(1)); // The "correct" phase that is going to use the predicted value and our measurement if(i%refresh == 0) { rectK = findROI(predictPt, detectedPedestrianFiltered); if(!fusionRects(prevRectK, rectK)) { cv::Point refPoint = rectCenter(rectK); // Get center point measurmt(0) = refPoint.x; measurmt(1) = refPoint.y; cv::Point measPt(measurmt(0),measurmt(1)); cv::Mat estim = Kalman.correct(measurmt); cv::Point statePt(estim.at<float>(0),estim.at<float>(1)); } prevRectK = rectK; } std::string str = std::to_string(sqrt(pow(Kalman.statePre.at<float>(2), 2)+pow(Kalman.statePre.at<float>(3), 2))); //cv::circle(sequence[i], measPt, 1, cv::Scalar(0,255,0), 7, 24); cv::circle(sequence[i], predictPt, 1, cv::Scalar(0,255,255), 7, 24); cv::putText(sequence[i], str, predictPt, cv::FONT_HERSHEY_PLAIN, 1.0, CV_RGB(255,0,0), 2.0); } cv::imshow("Video", sequence[i]); } //------------------CLEAR-VARIABLES------------------------------------// detectedPedestrian.clear(); featuresDetected.clear(); boxes.clear(); previousSequence.release(); flow.release(); imGray.release(); imGrayPrev.release(); roiHogDetected.clear(); backProj.clear(); //------------------CONDITIONS-ARRET-----------------------------------// if (cv::waitKey((int)(1000/fps)) == 27) //wait for 'esc' key press for 30ms. If 'esc' key is pressed, break loop { std::cout << "esc key is pressed by user" << std::endl; return 0; } } cv::waitKey(0); return 0; }
//カルマンフィルタ void kalmanFilter() { cv::VideoCapture capture(0); //x,y,vx,vy //cv::KalmanFilter kalman(4, 2, 0); cv::setIdentity(kalman.measurementMatrix, cv::Scalar(1.0)); cv::setIdentity(kalman.processNoiseCov, cv::Scalar::all(1e-5)); cv::setIdentity(kalman.measurementNoiseCov, cv::Scalar::all(0.1)); cv::setIdentity(kalman.errorCovPost, cv::Scalar::all(1.0)); // 等速直線運動モデル kalman.transitionMatrix = *(cv::Mat_<float>(4, 4) << 1,0,1,0, 0,1,0,1, 0,0,1,0, 0,0,0,1); //観測値を格納するもの cv::Mat_<float> measurement(2,1); measurement.setTo(cv::Scalar(0)); //初期値 kalman.statePre.at<float>(0) = initx; kalman.statePre.at<float>(1) = inity; kalman.statePre.at<float>(2) = 0; kalman.statePre.at<float>(3) = 0; cv::namedWindow("KalmanFilter", CV_WINDOW_AUTOSIZE); cv::setMouseCallback("KalmanFilter", _onMouse, 0); // メインループ while (!GetAsyncKeyState(VK_ESCAPE)) { // 画像を取得 capture >> capframe; cv::Mat image = capframe.clone(); // HSVに変換 //cv::Mat hsv = image.clone(); //cv::cvtColor(image, hsv, CV_BGR2HSV); std::cout << "HSV: (" << h << ", " << s << ", " << v << ")\n"; //クリックした点のRGBと近い画素を探索し、その重心を求める int sumX =0, sumY = 0, counter = 0; for(int y = 0; y < image.rows; y++) for(int x = 0; x < image.cols; x++) { //RGBの差分が閾値以内ならカウント if(sqrt((red - image.at<cv::Vec3b>(y,x)[2]) * (red - image.at<cv::Vec3b>(y,x)[2]) + (green - image.at<cv::Vec3b>(y,x)[1]) * (green - image.at<cv::Vec3b>(y,x)[1]) + (blue - image.at<cv::Vec3b>(y,x)[0]) * (blue - image.at<cv::Vec3b>(y,x)[0])) <= thresh) { //色付け image.at<cv::Vec3b>(y,x)[2] = 255; sumX += x; sumY += y; counter++; } } if(counter > 0) { //観測値 int mx = (int)(sumX/counter); int my = (int)(sumY/counter); measurement(0) = mx; measurement(1) = my; cv::Point measPt(measurement(0), measurement(1)); //表示 cv::circle(image, measPt, 10, cv::Scalar(0, 0, 255)); //修正フェーズ cv::Mat estimated = kalman.correct(measurement); cv::Point statePt(estimated.at<float>(0),estimated.at<float>(1)); } //予測フェーズ cv::Mat prediction = kalman.predict(); cv::Point predictPt(prediction.at<float>(0),prediction.at<float>(1)); // 表示 cv::circle(image, predictPt, 10, cv::Scalar(0, 255, 0)); cv::circle(image, cv::Point(20, 20), 10, CV_RGB(red, green, blue), 5); cv::putText(image,"target", cv::Point(0, 50), cv::FONT_HERSHEY_SIMPLEX, 0.7, CV_RGB(red, green, blue)); std::cout << (int)red << ", " << (int)green << ", " << (int)blue << std::endl; cv::imshow("KalmanFilter", image); cvWaitKey(1); } }