Detector::Status Detector::detectAndDraw()
{
const Scalar eye_color({0,255,255});
const Scalar face_color({255,128,0});
vector<Rect> objects;
vector<Rect> faces;
//check the data
if( eyesCascade.empty() || faceCascade.empty() )
{
return StatusNoModel;
}
image = imread(imageSrc.c_str() );
if (!image.data) {
return StatusNoImage;
}
if (!spectacles.data) {
return StatusNoSpectacles;
}
doThreshold(spectacles, specMask, 150, 255, CV_THRESH_BINARY_INV );
if (debug) {
namedWindow("Threshold");
imshow("Threshold",specMask);
}
//convert the image to grayscale
Mat gray(image.rows, image.cols, CV_8UC1);
cvtColor(image, gray, CV_BGR2GRAY);
equalizeHist(gray, gray);
// detect faces and eyes
faceCascade.detectMultiScale(gray, faces, 1.1, 2|CV_HAAR_SCALE_IMAGE);
if (debug) doMarkObjects(faces, face_color);
eyesCascade.detectMultiScale( gray, objects, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE );
if (debug) doMarkObjects(objects, eye_color);
//sort the results from left-to-right
std::sort(faces.begin(), faces.end(), by_x());
std::sort(objects.begin(), objects.end(), by_x());
//iterate over all found faces
pred_within_rect pwr;
for ( vector<Rect>::iterator face = faces.begin();face<faces.end();face++)
{
//process the detected face: if there are eyes found within it, then put spectacles on it
pwr.r = (*face);
vector<Rect>::iterator eye =std::find_if(objects.begin(), objects.end(),pwr);
if (eye!=objects.end())
{
doPutSpectaclesOnFace(*face, eye);
}
}
imshow(title, image);
return StatusOK;
}
//! 直方图均衡，为判断车牌颜色做准备
Mat CCharsSegment::histeq(Mat in)
{
Mat out(in.size(), in.type());
if(in.channels()==3)
{
Mat hsv;
vector<Mat> hsvSplit;
cvtColor(in, hsv, CV_BGR2HSV);
split(hsv, hsvSplit);
equalizeHist(hsvSplit[2], hsvSplit[2]);
merge(hsvSplit, hsv);
cvtColor(hsv, out, CV_HSV2BGR);
}
else if(in.channels()==1)
{
equalizeHist(in, out);
}
return out;
}
void
FaceDetector::run(){
CascadeClassifier cascade;
if (!cascade.load("res/haarcascade_frontalface_alt.xml") ) {
qDebug() << "Erreur lors du chargement du haar cascade frontalface...";
exit(0);
}
QSize processSize = keepAspectRatio(_image.cols, _image.rows, MAX_PROCESS_WIDTH, MAX_PROCESS_HEIGHT);
Mat color;
if ( _image.cols > processSize.width() ){
cv::resize( _image, color, cv::Size(processSize.width(), processSize.height() ) );
} else {
color = _image.clone();
}
QSize displaySize = keepAspectRatio(color.cols, color.rows);
_ratio = (double) color.cols/displaySize.width();
Mat gray;
cvtColor(color, gray, CV_BGR2GRAY);
//PRETRAITEMENT
equalizeHist(gray, gray);
vector<Rect> facesRects;
cascade.detectMultiScale(gray, facesRects, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, cvSize(16,16));
// qDebug() << "detected faces size: " << facesRects.size();
for (unsigned int i = 0; i < facesRects.size(); ++i){
Rect r = facesRects[i];
Mat croppedFace;
getRectSubPix(color, Size(r.width, r.height), Point2f(r.x + r.width/2, r.y + r.height / 2), croppedFace);
// qImageFaces << Mat2QImage(croppedFace);
struct DetectorData tmpData;
tmpData.id = i;
cv::resize( croppedFace, croppedFace, Size(WORK_SIZE,WORK_SIZE) );
tmpData.image = Mat2QImage(croppedFace);
emit sendFace( tmpData.image );
tmpData.rect = QRect(r.x/_ratio+1, r.y/_ratio+1, r.width/_ratio+1, r.height/_ratio+1);
tmpData.cvRect = r;
tmpData.mat = croppedFace;
cvtColor(croppedFace, tmpData.gray, CV_RGB2GRAY);
// cv::resize( tmpData.gray, tmpData.gray, Size(WORK_SIZE,WORK_SIZE) );
detectorData << tmpData;
}
qRegisterMetaType< QList<struct DetectorData> >( "QList<struct DetectorData>" );
emit detectionFinished( _index, detectorData );
}
cv::Mat FaceRecognizer::ImageNormalizer::normalize(const cv::Mat& in) const
{
// sanity check
if( in.type() != CV_8UC1 )
throw Util::Exception( UTIL_LOCSTRM << "face image must be grayscale (CV_8UC1)" );
// normalization
cv::medianBlur(in, tmp_[0], 1+2); // remove noise
equalizeHist(tmp_[0], tmp_[1]); // equalize histogram
cv::resize(tmp_[1], tmp_[0], size_); // resize image to the fixed-size
return tmp_[0];
}
/**
* @function detectAndDisplay
*/
int detectAndDisplay( cv::Mat frame ) {
std::vector<cv::Rect> faces;
//cv::Mat frame_gray;
std::vector<cv::Mat> rgbChannels(3);
cv::split(frame, rgbChannels);
cv::Mat frame_gray = rgbChannels[2];
face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE|CV_HAAR_FIND_BIGGEST_OBJECT, cv::Size(150, 150) );
// findSkin(debugImage);
for( int i = 0; i < faces.size(); i++ )
{
rectangle(debugImage, faces[i], 1234);
}
//-- Show what you got
if (faces.size() > 0) {
findEyes(frame_gray, faces[0]);
if (frame_gray.data) {
frame_gray.convertTo(frame_gray, CV_32FC1);
cv::Mat target = CropFace(frame_gray,leftPupil, rightPupil, offset_pct, dest_sz);
if (errorflag) {
return -1;
}
if (target.data) {
target.convertTo(target, CV_8UC1);
equalizeHist(target,target);
target.convertTo(target, CV_32FC1);
int index = query(target,ef);
index+= startnum;
char temp[3];
sprintf(temp, "%d", index);
std::string s(temp);
std::cout << "face recognized, index : " << index << std::endl;
// imshow("target", ef.norm_0_255(target).reshape(1, dest_sz.x));
imshow("result"+s, ef.getdb()[index-startnum]);
waitKey(0);
destroyWindow("result"+s);
return index;
}
}
}
return -1;
}
cv::Mat DetectRegions::histeq(const cv::Mat& in)
{
cv::Mat out(in.size(), in.type());
if (in.channels() == 3)
{
cv::Mat hsv;
std::vector<cv::Mat> hsvSplit;
cvtColor(in, hsv, CV_BGR2HSV);
split(hsv, hsvSplit);
equalizeHist(hsvSplit[2], hsvSplit[2]);
merge(hsvSplit, hsv);
cvtColor(hsv, out, CV_HSV2BGR);
}
else if (in.channels() == 1)
equalizeHist(in, out);
return out;
}
RawImage* OpenCvHistogramEqualization::ProcessInput(CommandLineArgModel* arg, RawImage* image)
{
OpenCvHistogramEqualizationModel* model = (OpenCvHistogramEqualizationModel*)arg->ParsedModel;
Rectangle roi = model->Roi;
Mat img = image->CloneToMat(roi);
vector<Mat> channels;
split(img, channels);
equalizeHist(channels[0], channels[0]);
equalizeHist(channels[1], channels[1]);
equalizeHist(channels[2], channels[2]);
merge(channels, img);
image->Import(img, roi.X, roi.Y);
return image;
}
void MainWindow::on_actionEqualize_triggered()
{
if(!activeMdiChild()) return;
CvGlWidget *actCV = activeMdiChild();
if(actCV->cvImage.channels() == 1){
equalizeHist(actCV->cvImage, actCV->cvImage);
actCV->showImage(actCV->cvImage);
actCV->buildHistogram();
makeHistogram();
}
}
/** @function detectAndDisplay */
void MotionDetection::FaceDetectByCascade(Mat frame, Mat mask, double ScaleFactor)
{
Mat frameCopy = frame.clone();
if(ScaleFactor != 1)
resize(frameCopy, frameCopy, Size(), ScaleFactor, ScaleFactor, INTER_AREA);
imshow("test1",frameCopy);
//initialize frameCopy_mask
// cvtColor(frameCopy_mask, frameCopy_mask, CV_BGR2GRAY);
vector<Rect> faces;
Mat frame_gray;
cvtColor( frameCopy, frame_gray, CV_BGR2GRAY );
//Apply histogram equalization with the function equalizeHist
equalizeHist( frame_gray, frame_gray );
face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(28, 28) );
Mat drawing = Mat::zeros(frame_gray.size(),CV_8UC1);
for( size_t i = 0; i < faces.size(); i++ )
{
//draw the face
//Point center( faces[i].x + faces[i].width*0.5, faces[i].y + faces[i].height*0.5 );
//ellipse( frame, center, Size( faces[i].width*0.5, faces[i].height*0.5), 0, 0, 360, Scalar( 255, 0, 255 ), 4, 8, 0 );
//slightly adjust the size of the face
faces[i] = Rect_AdjustSizeAroundCenter(faces[i],0.8,0.8);
// int tlx = faces[i].x, tly = faces[i].y;
// for(int j = tlx; j < (tlx + faces[i].width) ; j++)
// {
// for(int ii = tly; ii < (tly + faces[i].height); ii++)
// {
// frameCopy_mask.at<uchar>(Point(j,ii)) = 255;
// }
// }
rectangle( drawing, faces[i].tl(), faces[i].br(), Scalar(255,255,255), -1, 8, 0);
}
imshow("test2",drawing);
if(ScaleFactor != 1)
resize(drawing, drawing, Size(), 1/ScaleFactor, 1/ScaleFactor, INTER_NEAREST);
if(option_str == "-f")
{
static int counter = 0;
String output_str = outPut_Mask_Path + NumberToString(++counter) + ".png";
imwrite(output_str, drawing);
}
for(int i = 0; i < drawing.cols; i ++)
{
for(int j = 0; j < drawing.rows; j++)
{
Point p = Point(i,j);
if(drawing.at<uchar>(p) == 255)
mask.at<uchar>(p) += mask_add_step;
}
}
}
void equalize_fastnldenoising(vector<Mat>& frame)
{
float filter_length = 3;
int template_window_size = 7;
int search_window_size = 21;
for (int space = 0; space < frame.size(); space++)
{
//cvtColor(frame[space], frame[space], CV_BGR2GRAY);
equalizeHist(frame[space], frame[space]);
//fastNlMeansDenoising(frame[space], frame[space],template_window_size, search_window_size);
}
}
/** @function detectAndDisplay */
vector<PlateRegion> RegionDetector::doCascade(Mat frame)
{
//float scale_factor = 1;
int w = frame.size().width;
int h = frame.size().height;
vector<Rect> plates;
equalizeHist( frame, frame );
resize(frame, frame, Size(w * this->scale_factor, h * this->scale_factor));
//-- Detect plates
timespec startTime;
getTime(&startTime);
Size minSize(config->minPlateSizeWidthPx * this->scale_factor, config->minPlateSizeHeightPx * this->scale_factor);
Size maxSize(w * config->maxPlateWidthPercent * this->scale_factor, h * config->maxPlateHeightPercent * this->scale_factor);
if (config->opencl_enabled)
{
ocl::oclMat openclFrame(frame);
((ocl::OclCascadeClassifier*) plate_cascade)->detectMultiScale(openclFrame, plates, config->detection_iteration_increase, 3, 0, minSize, maxSize);
}
else
{
plate_cascade->detectMultiScale( frame, plates, config->detection_iteration_increase, 3,
0,
//0|CV_HAAR_SCALE_IMAGE,
minSize, maxSize );
}
if (config->debugTiming)
{
timespec endTime;
getTime(&endTime);
cout << "LBP Time: " << diffclock(startTime, endTime) << "ms." << endl;
}
for( int i = 0; i < plates.size(); i++ )
{
plates[i].x = plates[i].x / scale_factor;
plates[i].y = plates[i].y / scale_factor;
plates[i].width = plates[i].width / scale_factor;
plates[i].height = plates[i].height / scale_factor;
}
vector<PlateRegion> orderedRegions = aggregateRegions(plates);
return orderedRegions;
}
void ShapeFeature::compute(Mat& src)
{
resultingImage = src;
Mat channel[3];
// The actual splitting.
split(resultingImage, channel);
channel[1].setTo(Scalar(0));
Mat br;
merge(channel,3,br);
channel[0].setTo(Scalar(0));
Mat r;
merge(channel,3,r);
cvtColor(r, r, CV_BGR2GRAY);
medianBlur(r, r, 9 );
equalizeHist(r, r);
threshold(r, r, 120, 255, THRESH_BINARY);
erode(r, r, Mat());
dilate(r, r, Mat());
vector< vector <Point> > contours; // Vector for storing contour
vector< Vec4i > hierarchy;
int largest_contour_index=0;
int largest_area=0;
findContours( r.clone(), contours, hierarchy,RETR_EXTERNAL, CHAIN_APPROX_SIMPLE ); // Find the contours in the image
for( int i = 0; i< contours.size(); i++ ){
double a=contourArea( contours[i],false); // Find the area of contour
if(a>largest_area){
largest_area=a;
largest_contour_index=i; //Store the index of largest contour
}
}
Mat pointsf;
Mat(contours[largest_contour_index]).convertTo(pointsf, CV_32F);
RotatedRect boxE = fitEllipse(pointsf);
if(boxE.size.width<boxE.size.height)
value = (float)boxE.size.width/boxE.size.height;
else
value = (float)boxE.size.height/boxE.size.width;
}
void FaceDetector::findFacesInImage(const Mat *img, vector<Rect> *res) {
Mat tmp;
// chuyển đổi hình ảnh sang màu xám và bình thường hóa biểu đồ:
cvtColor(*img, tmp, CV_BGR2GRAY);
//Cân bằng lại
equalizeHist(tmp, tmp);
// làm sạch vector
res->clear();
// phát hiện khuôn mặt:
_cascade->detectMultiScale(tmp, *res, DET_SCALE_FACTOR, DET_MIN_NEIGHBORS, CV_HAAR_FIND_BIGGEST_OBJECT | CV_HAAR_SCALE_IMAGE, Size(30, 30));
}
//get position of center of eye
CvPoint2D32f getEyeCenter(Mat &face_frame, Rect eye){
//crop eye to get rid of some noise
int crop_y = 10;
eye.y += crop_y;
eye.height -= crop_y * 2;
//eye.width -= 20;
Mat eye_color = face_frame(eye);
Mat eye_scelra;
//convert to HSV and get saturation channel
cvtColor(eye_color, eye_scelra, CV_RGB2HSV);
std::vector<Mat> hsvChannels(3);
split(eye_scelra, hsvChannels);
eye_scelra = hsvChannels[1].mul(hsvChannels[2]) / 30;
//invert
bitwise_not(eye_scelra, eye_scelra);
//blur
blur(eye_scelra, eye_scelra, Size(4,4));
//apply histogram equalization
equalizeHist(eye_scelra, eye_scelra);
//threshold
//threshold(eye_scelra, eye_scelra, 10, 255, THRESH_BINARY_INV); //threshold type 3, thesh. to 0
//calc center of mass
float sum = 0;
float sum_x = 0;
float sum_y = 0;
for (int y = 0; y < eye_scelra.rows; ++y)
{
uchar* row = eye_scelra.ptr<uchar>(y);
for (int x = 0; x < eye_scelra.cols; ++x)
{
sum += row[x];
sum_x += row[x]*x;
sum_y += row[x]*y;
}
}
CvPoint2D32f max = cvPoint2D32f(sum_x/sum, sum_y/sum);
//circle(eye_scelra, max, 3, 0);
//imshow("eye", eye_scelra);
//adjust for crop
max.y += crop_y;
return max;
}
bool OpenCVItemProcessing::Loop()
{
char exit = waitKey(20);
if (exit == 27) return false;
else
{
if (camera.isOpened())
{
//camera.set(CV_CAP_PROP_EXPOSURE, (double)exposure - 11.0);
//Grab our image
camera.read(videoImage);
std::vector<Mat> videoCopy;
split(videoImage, videoCopy);
cvtColor(videoImage, videoImage, CV_BGR2GRAY);
std::vector<Mat> channels;
split(videoImage, channels);
equalizeHist(channels[0], channels[0]);
merge(channels, videoImage);
cvtColor(videoImage, videoImage, CV_GRAY2BGR);
//Make a pseudo grayscale image
//ImageProcess_ColorMask(videoImage);
//actually convert to grayscale
//cvtColor(videoImage, videoImage, CV_BGR2GRAY);
//To Binary
//threshold(videoImage, videoImage, 25.0, 255.0, THRESH_BINARY);
//Canny?
//Canny(videoImage, videoImage, 0.0, 1.0, 3, false);
//Show image on screen
imshow(windowName, videoImage);
}
return true;
}
}
/*
* Returns a list of all the vehicles found in a frame.
* The base classifier used for this process can be set in CLASSIFIER_PATH.
*/
vector<Rect> RoadDetection::getVehicles(Mat frame)
{
Mat copy, equalizedFrame;
CascadeClassifier vehiclesCascade;
vector<Rect> vehicles;
frame.copyTo(copy);
cvtColor(copy, equalizedFrame, CV_BGR2GRAY);
equalizeHist(equalizedFrame, equalizedFrame);
vehiclesCascade.load(CLASSIFIER_PATH);
vehiclesCascade.detectMultiScale(equalizedFrame, vehicles, CASCADE_SCALE, CASCADE_MIN_NEIGHBORS, 0 | CASCADE_SCALE_IMAGE, Size(CASCADE_MIN_SIZE, CASCADE_MAX_SIZE));
return vehicles;
}
void FaceDetectAnalytic::process(analytic::ConcurrentQueue<analytic::api::Image_t>* pInputQueue, analytic::ConcurrentQueue<analytic::api::Image_t>* pOutputQueue)
{
/* Analytic process starts from here */
while(pHaarCascade)
{
/* 1. get a image from input queue */
api::Image_t image = pInputQueue->pop();
cv::Mat matInputImage = image.matImage; // got it
/* 2. clone the input image */
cv::Mat matToBeProcessed = matInputImage.clone();
/* 3. release the input image */
matInputImage.release(); //please do this release!
/* 4. OK, now use cloned image for analysis */
// Start processing
cv::Mat matGray;
cv::cvtColor(matToBeProcessed, matGray, CV_BGR2GRAY);
equalizeHist(matGray, matGray);
vector<Rect_<int> > vFaces;
pHaarCascade->detectMultiScale(matGray, vFaces, _dScaleFactor, _iMinNeighbors, 0|CV_HAAR_SCALE_IMAGE, _minSize, _maxSize);
for (size_t i = 0; i < vFaces.size(); ++i) {
Rect recFace = vFaces[i];
Mat matFace = matGray(recFace);
rectangle(matToBeProcessed, recFace, CV_RGB(0, 255, 0), 1);
matFace.release();
}
if(_bDisplayOutput)
{
cv::imshow("Output", matToBeProcessed);
cv::waitKey(1);
}
// freeing generated Mats
matGray.release();
matToBeProcessed.release();
/* 5. set output */
if(vFaces.size() > 0)
{
image.bGenerateAnalyticEvent = true;
resultXml(vFaces, image.sCustomTextResult);
}
/* 6. push into output queue */
pOutputQueue->push(image);
}
}
void Detection::getFaces(Mat& a) {
std::vector<Rect> faces;
Mat frameGray;
Mat frameGray2;
cvtColor( a, frameGray, CV_BGR2GRAY );
equalizeHist( frameGray, frameGray2 );
face_cascade.detectMultiScale( frameGray2, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );
for( int i = 0; i < faces.size(); i++ )
{
rectangle(a, faces[i], Scalar(255, 0, 0));
}
std::cout << "end loop "<< std::endl;
}
/** @function detectAndDisplay */
vector<Rect> RegionDetector::doCascade(Mat frame)
{
//float scale_factor = 1;
int w = frame.size().width;
int h = frame.size().height;
vector<Rect> plates;
equalizeHist( frame, frame );
resize(frame, frame, Size(w * this->scale_factor, h * this->scale_factor));
//-- Detect plates
timespec startTime;
getTime(&startTime);
Size minSize(config->minPlateSizeWidthPx * this->scale_factor, config->minPlateSizeHeightPx * this->scale_factor);
Size maxSize(w * config->maxPlateWidthPercent * this->scale_factor, h * config->maxPlateHeightPercent * this->scale_factor);
plate_cascade.detectMultiScale( frame, plates, 1.1, 3,
0,
//0|CV_HAAR_SCALE_IMAGE,
minSize, maxSize );
if (config->debugTiming)
{
timespec endTime;
getTime(&endTime);
cout << "LBP Time: " << diffclock(startTime, endTime) << "ms." << endl;
}
for( int i = 0; i < plates.size(); i++ )
{
plates[i].x = plates[i].x / scale_factor;
plates[i].y = plates[i].y / scale_factor;
plates[i].width = plates[i].width / scale_factor;
plates[i].height = plates[i].height / scale_factor;
}
return plates;
}
void caasCLR4Tx1::FindIsolatorAngle()
{
//We cut 4/5 width of isolator out
int height = 3 * (isolatorBottomEdge - isolatorTopEdge) / 2; int middle = (isolatorBottomEdge + isolatorTopEdge) / 2;
Rect rect = Rect(isolatorRightEdge - 4 * isolatorWidth / 5, middle - height / 2, 4 * isolatorWidth / 5, height);
Mat imageIsolator = imageGray(rect);
int scale = 4;
resize(imageIsolator, imageIsolator, Size(imageIsolator.cols / scale, imageIsolator.rows / scale));
#if _DEBUG
imwrite("Isolator.jpg", imageIsolator);
#endif
//sharpen the image
//Unsharping masking: Use a Gaussian smoothing filter and subtract the smoothed version from the original image (in a weighted way so the values of a constant area remain constant).
Mat imageBlurred, imageGraySharpened; double GAUSSIAN_RADIUS = 4.0;
GaussianBlur(imageIsolator, imageBlurred, Size(0, 0), GAUSSIAN_RADIUS);
addWeighted(imageIsolator, 1.5, imageBlurred, -0.5, 0, imageGraySharpened);
#if _DEBUG
//imageGraySharpened = imageGrayQuarter;
imwrite("IsolatorSharpened.jpg", imageGraySharpened);
#endif
imageGraySharpened = imageIsolator;
//Histogram Equalization
equalizeHist(imageGraySharpened, imageGraySharpened);
#if _DEBUG
imwrite("IsolatorEqualized.jpg", imageGraySharpened);
#endif
//Otsu binarization
Mat imageOtsu; threshold(imageGraySharpened, imageOtsu, 0, 255, CV_THRESH_BINARY | CV_THRESH_OTSU);
#if _DEBUG
imwrite("IsolatorOtsu.jpg", imageOtsu);
#endif
//Canny Edge Detection
int median = Median(imageGraySharpened);
Mat imageCanny; Canny(imageGraySharpened, imageCanny, 0.66 * median, 1.33 * median);
#if _DEBUG
imwrite("IsolatorCanny.jpg", imageCanny);
#endif
}
void CSkin::face_detect(const cv::Mat &frame)
{
char *face_cascade_name = "haarcascade_frontalface_alt.xml";
cv::CascadeClassifier face_cascade;
if (!face_cascade.load(face_cascade_name))
{
return;
}
if (frame.cols > 0 && frame.rows > 0)
{
cv::Mat frame_gray;
if (frame.channels() == 3)
cvtColor(frame, frame_gray, CV_BGR2GRAY);
else
frame_gray = frame.clone();
equalizeHist(frame_gray, frame_gray);
face_cascade.detectMultiScale(frame_gray, faces, 1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, cv::Size(30, 30));
}
}
void FaceDetect::detect(cv::Mat& img, fd_cb_t fd_cb) {
int i = 0;
double t = 0;
std::vector<cv::Rect> faces;
const static cv::Scalar colors[] = { CV_RGB(0,0,255),
CV_RGB(0,128,255),
CV_RGB(0,255,255),
CV_RGB(0,255,0),
CV_RGB(255,128,0),
CV_RGB(255,255,0),
CV_RGB(255,0,0),
CV_RGB(255,0,255)} ;
cv::Mat gray, smallImg( cvRound (img.rows/scale_), cvRound(img.cols/scale_), CV_8UC1 );
cv::cvtColor( img, gray, CV_BGR2GRAY );
cv::resize( gray, smallImg, smallImg.size(), 0, 0, cv::INTER_LINEAR );
equalizeHist( smallImg, smallImg );
t = (double)cvGetTickCount();
cascade_.detectMultiScale( smallImg, faces,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
|CV_HAAR_SCALE_IMAGE
,
cv::Size(30, 30) );
fd_cb_t cb = fd_cb ? fd_cb : fd_cb_;
for( std::vector<cv::Rect>::const_iterator r = faces.begin(); r != faces.end(); r++, i++ )
{
cv::Mat smallImgROI;
cv::Point center;
cv::Scalar color = colors[i%8];
int radius;
center.x = cvRound((r->x + r->width*0.5)*scale_);
center.y = cvRound((r->y + r->height*0.5)*scale_);
radius = cvRound((r->width + r->height)*0.25*scale_);
if(cb) {
cb(center.x, center.y, radius);
}
}
}
vector<DetectedObject> CascadeObjectDetector::detectObjects(Mat image) {
Mat gray;
if (image.channels() > 1) {
cvtColor(image, gray, CV_RGB2GRAY);
} else {
gray = image;
}
equalizeHist(gray, gray);
detector->process(gray);
vector<Rect> found;
detector->getObjects(found);
vector<DetectedObject> objects;
for (int i = 0; i < found.size(); i++) {
DetectedObject o;
o.boundingBox = found[i];
objects.push_back(o);
}
return objects;
}
//method to detect faces in the image frame and return rectangular ROI if found
Rect detect(Mat &frame)
{
int flag=0;
Mat f1(240/2,320/2,frame.type ());
cv::flip(frame,frame,1);
cv::resize(frame,f1,f1.size(),0,0,INTER_CUBIC);
//cv::flip(f1,f1,1);
//ROI of face locations
std::vector<Rect> faces;
Mat frame_gray;
//converting to grayscale
cvtColor( f1, frame_gray, CV_BGR2GRAY );
//pre processing frame using histogram equalization
equalizeHist( frame_gray, frame_gray );
//-- multiscale detection of faces
face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );
float maxarea=0;
Rect rect;
for( int i = 0; i < faces.size(); i++ )
{
Point center( faces[i].x + faces[i].width*0.5, faces[i].y + faces[i].height*0.5 );
int scalex=frame.cols/f1.cols;
int scaley=frame.rows/f1.rows;
Point ncenter;
ncenter.x=center.x*scalex;
ncenter.y=center.y*scaley;
Point corner;
corner.x=faces[i].x;
corner.y=faces[i].y;
center.x=corner.x*scalex;
center.y=corner.y*scaley;
if(faces[i].width*faces[i].height>maxarea)
{
maxarea=faces[i].width*faces[i].height;
//scale the center and dimension to original image scale
rect=Rect(center.x,center.y,faces[i].width*scalex,faces[i].height*scaley);
//rect=Rect(faces[i].x,faces[i].y,faces[i].width,faces[i].height);
}
}
return rect;
}
void Detection::getFaces(Mat& a) {
std::vector<Rect> faces;
Mat frameGray;
cvtColor( a, frameGray, CV_BGR2GRAY );
equalizeHist( frameGray, frameGray );
//-- Detect faces
face_cascade.detectMultiScale( frameGray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );
std::cout << "begin loop "<< std::endl;
for( int i = 0; i < faces.size(); i++ )
{
Point center( faces[i].x + faces[i].width*0.5, faces[i].y + faces[i].height*0.5 );
rectangle(a, faces[i], Scalar(255, 0, 0));
//ellipse( a, center, Size( faces[i].width*0.5, faces[i].height*0.5), 0, 0, 360, Scalar( 255, 0, 255 ), 4, 8, 0 );
std::cout << "iteration " + i << std::endl;
}
std::cout << "end loop "<< std::endl;
}
vector<RotatedRect> grayscale_method(Mat& image, Mat& imgMask) {
// std::cout << "SAMPLE_DETECTOR::Starting Grayscale METHOD" << std::endl;
cvtColor(image, grayscale_image, CV_BGR2GRAY);
Mat eq_gray;
equalizeHist(grayscale_image,eq_gray);
cv::imwrite("Sample-04-Grayscale.png", grayscale_image);
cv::imwrite("Sample-Grayscale-equalized.png", eq_gray);
threshold(eq_gray, grayscale_filtered_image, min_grayscale_thresh, max_grayscale_thresh, 0);
cv::imwrite("Sample-05-Grayscale-Threshold.png", grayscale_filtered_image);
// Erode and Dilate
obj_tracker.filter(grayscale_filtered_image, grayscale_erode_size, grayscale_dilate_size);
cv::imwrite("Sample-06-Grayscale-Filtered.png", grayscale_filtered_image);
// std::cout << "SAMPLE_DETECTOR::Completed Grayscale METHOD" << std::endl;
return obj_tracker.track(image,grayscale_filtered_image,imgMask);
}
Mat equalizeIntensity(const Mat& inputImage)
{
if(inputImage.channels() >= 3)
{
Mat ycrcb;
cvtColor(inputImage,ycrcb,CV_BGR2YCrCb);
vector<Mat> channels;
split(ycrcb,channels);
equalizeHist(channels[0], channels[0]);
Mat result;
merge(channels,ycrcb);
cvtColor(ycrcb,result,CV_YCrCb2BGR);
return result;
}
return Mat();
}
Mat eql(Mat img, Mat dst) {
if(img.data < 0){
cout << "Not loaded" << endl;
}
Mat image = img.clone();
int histSize = 256;
float range[] = { 0, 256 } ;
const float* ranges[] = { range };
MatND hist;
calcHist( &image, 1, 0, Mat(), hist, 1, &histSize, ranges, true, false );
// Show the calculated histogram in command window
//double total;
//total = image.rows * image.cols;
equalizeHist(image,dst);
}
/********************************************************************
Utils::HistogramEqualize
Performs Histogram Equalization on the given image
Exceptions:
None
*********************************************************************/
cv::Mat Utils::HistogramEqualize(const cv::Mat& inputImage)
{
if ( inputImage.channels() >= 3 )
{
cv::Mat ycrcb;
cvtColor(inputImage,ycrcb,CV_BGR2YCrCb);
std::vector<cv::Mat> channels;
split(ycrcb,channels);
equalizeHist(channels[0], channels[0]);
cv::Mat result;
merge(channels,ycrcb);
cvtColor(ycrcb,result,CV_YCrCb2BGR);
return result;
}
return cv::Mat();
}
////////////////////////////////////////////////////////////
// Helper Function for Cascade Classifier
// Note: Cascade classifier is not currently used
// by this application. We left the functions associated
// with this method for future reference.
////////////////////////////////////////////////////////////
void detectAndDisplay(Mat image, string panel_cascade_name)
{
CascadeClassifier panel_cascade;
if (!panel_cascade.load(panel_cascade_name)){ printf("--(!)Error loading\n"); return; };
std::vector<Rect> detectedPanels;
Mat frame_gray;
cvtColor(image, frame_gray, CV_BGR2GRAY);
equalizeHist(frame_gray, frame_gray);
//-- Detect faces
panel_cascade.detectMultiScale(frame_gray, detectedPanels, 1.1, 20, 0 | CV_HAAR_SCALE_IMAGE, Size(200, 200));
for (size_t i = 0; i < detectedPanels.size(); i++)
{
Point topLeft(detectedPanels[i].x, detectedPanels[i].y);
Point botRight(detectedPanels[i].x + detectedPanels[i].width, detectedPanels[i].y + detectedPanels[i].height);
rectangle(image, topLeft, botRight, Scalar(0, 0, 255), 4);
}
//-- Show what you got
imshow("Classifier Result", image);
}
