Basic Haar cascade/CamShift tracking.

interface/src/Webcam.cpp

@@ -19,9 +19,11 @@
 #include "Webcam.h"
 
 using namespace cv;
+using namespace std;
 
 // register OpenCV matrix type with Qt metatype system
 int matMetaType = qRegisterMetaType<Mat>("cv::Mat");
+int rotatedRectMetaType = qRegisterMetaType<RotatedRect>("cv::RotatedRect");
 
 Webcam::Webcam() : _enabled(false), _frameTextureID(0)  {
     // the grabber simply runs as fast as possible
@@ -69,6 +71,17 @@ void Webcam::renderPreview(int screenWidth, int screenHeight) {
         glBindTexture(GL_TEXTURE_2D, 0);
         glDisable(GL_TEXTURE_2D);
         
+        glBegin(GL_LINE_LOOP);
+            Point2f facePoints[4];
+            _faceRect.points(facePoints);
+            float xScale = previewWidth / (float)_frameWidth;
+            float yScale = PREVIEW_HEIGHT / (float)_frameHeight;
+            glVertex2f(left + facePoints[0].x * xScale, top + facePoints[0].y * yScale);
+            glVertex2f(left + facePoints[1].x * xScale, top + facePoints[1].y * yScale);
+            glVertex2f(left + facePoints[2].x * xScale, top + facePoints[2].y * yScale);
+            glVertex2f(left + facePoints[3].x * xScale, top + facePoints[3].y * yScale);
+        glEnd();
+        
         char fps[20];
         sprintf(fps, "FPS: %d", (int)(roundf(_frameCount * 1000000.0f / (usecTimestampNow() - _startTimestamp))));
         drawtext(left, top + PREVIEW_HEIGHT + 20, 0.10, 0, 1, 0, fps);
@@ -83,25 +96,26 @@ Webcam::~Webcam() {
     delete _grabber;
 }
 
-void Webcam::setFrame(const Mat& frame) {
+void Webcam::setFrame(const Mat& frame, const RotatedRect& faceRect) {
     IplImage image = frame;
-    glPixelStorei(GL_UNPACK_ROW_LENGTH, image.widthStep);
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, image.widthStep / 3);
     if (_frameTextureID == 0) {
         glGenTextures(1, &_frameTextureID);
         glBindTexture(GL_TEXTURE_2D, _frameTextureID);
-        glTexImage2D(GL_TEXTURE_2D, 0, GL_LUMINANCE, _frameWidth = image.width, _frameHeight = image.height, 0, GL_LUMINANCE,
+        glTexImage2D(GL_TEXTURE_2D, 0, GL_RGB, _frameWidth = image.width, _frameHeight = image.height, 0, GL_BGR,
             GL_UNSIGNED_BYTE, image.imageData);
         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
         printLog("Capturing webcam at %dx%d.\n", _frameWidth, _frameHeight);
     
     } else {
         glBindTexture(GL_TEXTURE_2D, _frameTextureID);
-        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, _frameWidth, _frameHeight, GL_LUMINANCE, GL_UNSIGNED_BYTE, image.imageData);    
+        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, _frameWidth, _frameHeight, GL_BGR, GL_UNSIGNED_BYTE, image.imageData);    
     }
     glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
     glBindTexture(GL_TEXTURE_2D, 0);
     
-    // update our frame count for fps computation
+    // store our face rect, update our frame count for fps computation
+    _faceRect = faceRect;
     _frameCount++;
     
     const int MAX_FPS = 60;
@@ -119,6 +133,9 @@ void Webcam::setFrame(const Mat& frame) {
     QTimer::singleShot(qMax((int)remaining / 1000, 0), _grabber, SLOT(grabFrame()));
 }
 
+FrameGrabber::FrameGrabber() : _capture(0), _searchWindow(0, 0, 0, 0) {
+}
+
 FrameGrabber::~FrameGrabber() {
     if (_capture != 0) {
         cvReleaseCapture(&_capture);
@@ -139,6 +156,11 @@ void FrameGrabber::grabFrame() {
 #ifdef __APPLE__
         configureCamera(0x5ac, 0x8510, false, 0.99, 0.5, 0.5, 0.5, true, 0.5);
 #endif
+
+        switchToResourcesParentIfRequired();
+        if (!_faceCascade.load("resources/haarcascades/haarcascade_frontalface_alt.xml")) {
+            printLog("Failed to load Haar cascade for face tracking.\n");
+        }
     }
     IplImage* image = cvQueryFrame(_capture);
     if (image == 0) {
@@ -153,10 +175,42 @@ void FrameGrabber::grabFrame() {
         return;
     }
     
-    // convert to grayscale and equalize
+    // if we don't have a search window (yet), try using the face cascade
-    Mat frame = image, grayFrame;
+    Mat frame = image;
-    cvtColor(frame, grayFrame, CV_BGR2GRAY);
+    int channels = 0;
-    equalizeHist(grayFrame, grayFrame);
+    float ranges[] = { 0, 180 };
-    
+    const float* range = ranges;
-    QMetaObject::invokeMethod(Application::getInstance()->getWebcam(), "setFrame", Q_ARG(cv::Mat, grayFrame));
+    if (_searchWindow.area() == 0) {
+        vector<Rect> faces;
+        _faceCascade.detectMultiScale(frame, faces, 1.1, 6);
+        if (!faces.empty()) {
+            _searchWindow = faces.front();
+            updateHSVFrame(frame);
+        
+            Mat faceHsv(_hsvFrame, _searchWindow);
+            Mat faceMask(_mask, _searchWindow);
+            int sizes = 30;
+            calcHist(&faceHsv, 1, &channels, faceMask, _histogram, 1, &sizes, &range);
+            double min, max;
+            minMaxLoc(_histogram, &min, &max);
+            _histogram.convertTo(_histogram, -1, (max == 0.0) ? 0.0 : 255.0 / max);
+        }
+    }
+    RotatedRect faceRect;
+    if (_searchWindow.area() > 0) {
+        updateHSVFrame(frame);
+        
+        calcBackProject(&_hsvFrame, 1, &channels, _histogram, _backProject, &range);
+        bitwise_and(_backProject, _mask, _backProject);
+        
+        faceRect = CamShift(_backProject, _searchWindow, TermCriteria(CV_TERMCRIT_EPS | CV_TERMCRIT_ITER, 10, 1));
+        _searchWindow = faceRect.boundingRect();
+    }   
+    QMetaObject::invokeMethod(Application::getInstance()->getWebcam(), "setFrame",
+        Q_ARG(cv::Mat, frame), Q_ARG(cv::RotatedRect, faceRect));
+}
+
+void FrameGrabber::updateHSVFrame(const Mat& frame) {
+    cvtColor(frame, _hsvFrame, CV_BGR2HSV);
+    inRange(_hsvFrame, Scalar(0, 55, 65), Scalar(180, 256, 256), _mask);
 }

interface/src/Webcam.h

@@ -36,7 +36,7 @@ public:
 public slots:
     
     void setEnabled(bool enabled);
-    void setFrame(const cv::Mat& image);
+    void setFrame(const cv::Mat& image, const cv::RotatedRect& faceRect);
     
 private:
     
@@ -47,6 +47,7 @@ private:
     int _frameWidth;
     int _frameHeight;
     GLuint _frameTextureID;
+    cv::RotatedRect _faceRect;
     
     long long _startTimestamp;
     int _frameCount;
@@ -59,7 +60,7 @@ class FrameGrabber : public QObject {
     
 public:
     
-    FrameGrabber() : _capture(0) { }
+    FrameGrabber();
     virtual ~FrameGrabber();
 
 public slots:
@@ -68,10 +69,18 @@ public slots:
     
 private:
     
+    void updateHSVFrame(const cv::Mat& frame);
+    
     CvCapture* _capture;
     cv::CascadeClassifier _faceCascade;
+    cv::Mat _hsvFrame;
+    cv::Mat _mask;
+    cv::SparseMat _histogram;
+    cv::Mat _backProject;
+    cv::Rect _searchWindow;
 };
 
 Q_DECLARE_METATYPE(cv::Mat)
+Q_DECLARE_METATYPE(cv::RotatedRect)
 
 #endif /* defined(__interface__Webcam__) */