/// Testing cartToDisp and dispToCart
bool testSparsePointConversion() {
CvStereoCamModel camModel;
camModel.setCameraParams(FX, FY, TX, CLX, CRX, CY, DISPUNITSCALE);
char filename[]= "test/pointsInCartesian.xml";
CvMat* points_cart = (CvMat*)cvLoad(filename);
if (points_cart == NULL ) {
printf("Cannot load data file: %s\n", filename);
return false;
}
CvMat* points_cart0 = (CvMat*)cvCreateMat(points_cart->rows, points_cart->cols,
CV_64FC1);
CvMat* points_disp = (CvMat*)cvCreateMat(points_cart->rows, points_cart->cols,
CV_64FC1);
CvMat* points_diff = (CvMat*)cvCreateMat(points_cart->rows, points_cart->cols,
CV_64FC1);
camModel.cartToDisp(points_cart, points_disp);
camModel.dispToCart(points_disp, points_cart0);
cvAbsDiff(points_cart, points_cart0, points_diff);
double min_val, max_val;
cvMinMaxLoc(points_diff, &min_val, &max_val);
if (max_val > .1e-10) {
printf("dispToCart() or cartToDisp() failed: max diff = %f\n", max_val);
return false;
}
cvReleaseMat(&points_cart);
cvReleaseMat(&points_disp);
cvReleaseMat(&points_diff);
return true;
}
/// The image callback. For each new image, copy it, perform face detection, and draw the rectangles on the image.
void image_cb_all(ros::Time t)
{
detect_++;
if (detect_ % 2) {
return;
}
cv_mutex_.lock();
CvSize im_size;
// Convert the images to OpenCV format
if (lbridge_.fromImage(limage_,"bgr")) {
cv_image_left_ = lbridge_.toIpl();
}
if (dbridge_.fromImage(dimage_)) {
cv_image_disp_ = dbridge_.toIpl();
}
// Convert the stereo calibration into a camera model.
if (cam_model_) {
delete cam_model_;
}
double Fx = rcinfo_.P[0];
double Fy = rcinfo_.P[5];
double Clx = rcinfo_.P[2];
double Crx = Clx;
double Cy = rcinfo_.P[6];
double Tx = -rcinfo_.P[3]/Fx;
cam_model_ = new CvStereoCamModel(Fx,Fy,Tx,Clx,Crx,Cy,1.0/(double)stinfo_.dpp);
if ( cv_image_left_ ) {
im_size = cvGetSize(cv_image_left_);
vector<CvRect> faces_vector = people_->detectAllFaces(cv_image_left_, haar_filename_, 1.0, cv_image_disp_, cam_model_, true);
// Get the average disparity in the middle half of the bounding box, and compute the face center in 3d.
// Publish the face center as a track point.
if (cv_image_disp_ && cam_model_) {
int r, c, good_pix;
ushort* ptr;
double avg_disp;
CvRect *one_face;
robot_msgs::PositionMeasurement pos;
CvMat *uvd = cvCreateMat(1,3,CV_32FC1);
CvMat *xyz = cvCreateMat(1,3,CV_32FC1);
for (uint iface = 0; iface < faces_vector.size(); iface++) {
one_face = &faces_vector[iface];
good_pix = 0;
avg_disp = 0;
for (r = floor(one_face->y+0.25*one_face->height); r < floor(one_face->y+0.75*one_face->height); r++) {
ptr = (ushort*)(cv_image_disp_->imageData + r*cv_image_disp_->widthStep);
for (c = floor(one_face->x+0.25*one_face->width); c < floor(one_face->x+0.75*one_face->width); c++) {
if (ptr[c] > 0) {
avg_disp += ptr[c];
good_pix++;
}
}
}
avg_disp /= (double)good_pix; // Take the average.
cvmSet(uvd,0,0,one_face->x+one_face->width/2.0);
cvmSet(uvd,0,1,one_face->y+one_face->height/2.0);
cvmSet(uvd,0,2,avg_disp);
cam_model_->dispToCart(uvd,xyz);
bool do_publish = true;
if (external_init_) {
// Check if this person's face is close enough to one of the previously known faces and associate it.
// If not, don't publish it.
}
if (do_publish) {
pos.header.stamp = limage_.header.stamp;
pos.name = "face_detection";
pos.object_id = "-1";
pos.pos.x = cvmGet(xyz,0,2);
pos.pos.y = -1.0*cvmGet(xyz,0,0);
pos.pos.z = -1.0*cvmGet(xyz,0,1);
pos.header.frame_id = "stereo_link";
pos.reliability = 0.8;
pos.initialization = 0;
//pos.covariance = ;
publish("person_measurement",pos);
}
}
cvReleaseMat(&uvd);
cvReleaseMat(&xyz);
if (do_display_) {
if (!cv_image_disp_out_) {
cv_image_disp_out_ = cvCreateImage(im_size,IPL_DEPTH_8U,1);
}
cvCvtScale(cv_image_disp_,cv_image_disp_out_,4.0/stinfo_.dpp);
cvShowImage("Face detector: Face Detection",cv_image_left_);
cvShowImage("Face detector: Disparity",cv_image_disp_out_);
}
}
}
cv_mutex_.unlock();
}