I have a problem in my iOS application where i attempt to obtain a view matrix using solvePnP and render a 3d cube using modern OpenGL. While my code attempts to render a 3d cube directly on top of the detected marker, it seems to render with a certain offset :
(on the bottom right of the image you can see an opencv render of the homography around the tracker marker. the rest of the screen is an opengl render of the camera input frame and a 3d cube at location (0,0,0).
the cube rotates and translates correctly whenever i move the marker, though it is very telling that there is some difference in the scale of translations (IE, if i move my marker 5cm in the real world, it hardly moves by 1cm on screen)
these are what i believe to be the relevant parts of the code where the error could come from :
Extracting view matrix from homography :
AVCaptureDevice *deviceInput = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
AVCaptureDeviceFormat *format = deviceInput.activeFormat;
CMFormatDescriptionRef fDesc = format.formatDescription;
CGSize dim = CMVideoFormatDescriptionGetPresentationDimensions(fDesc, true, true);
const float cx = float(dim.width) / 2.0;
const float cy = float(dim.height) / 2.0;
const float HFOV = format.videoFieldOfView;
const float VFOV = ((HFOV)/cx)*cy;
const float fx = abs(float(dim.width) / (2 * tan(HFOV / 180 * float(M_PI) / 2)));
const float fy = abs(float(dim.height) / (2 * tan(VFOV / 180 * float(M_PI) / 2)));
Mat camIntrinsic = Mat::zeros(3, 3, CV_64F);
camIntrinsic.at<double>(0, 0) = fx;
camIntrinsic.at<double>(0, 2) = cx;
camIntrinsic.at<double>(1, 1) = fy;
camIntrinsic.at<double>(1, 2) = cy;
camIntrinsic.at<double>(2, 2) = 1.0;
std::vector<cv::Point3f> object3dPoints;
object3dPoints.push_back(cv::Point3f(-0.5f,-0.5f,0));
object3dPoints.push_back(cv::Point3f(+0.5f,-0.5f,0));
object3dPoints.push_back(cv::Point3f(+0.5f,+0.5f,0));
object3dPoints.push_back(cv::Point3f(-0.5f,+0.5f,0));
cv::Mat raux,taux;
cv::Mat Rvec, Tvec;
cv::solvePnP(object3dPoints, mNewImageBounds, camIntrinsic, Mat(),raux,taux); //mNewImageBounds are the 4 corner of the homography detected by perspectiveTransform (the green outline seen in the image)
raux.convertTo(Rvec,CV_32F);
taux.convertTo(Tvec ,CV_64F);
Mat Rot(3,3,CV_32FC1);
Rodrigues(Rvec, Rot);
// [R | t] matrix
Mat_<double> para = Mat_<double>::eye(4,4);
Rot.convertTo(para(cv::Rect(0,0,3,3)),CV_64F);
Tvec.copyTo(para(cv::Rect(3,0,1,3)));
Mat cvToGl = Mat::zeros(4, 4, CV_64F);
cvToGl.at<double>(0, 0) = 1.0f;
cvToGl.at<double>(1, 1) = -1.0f; // Invert the y axis
cvToGl.at<double>(2, 2) = -1.0f; // invert the z axis
cvToGl.at<double>(3, 3) = 1.0f;
para = cvToGl * para;
Mat_<double> modelview_matrix;
Mat(para.t()).copyTo(modelview_matrix); // transpose to col-major for OpenGL
for(int col = 0; col < modelview_matrix.cols; col++)
{
for(int row = 0; row < modelview_matrix.rows; row++)
{
openGLViewMatrix[col][row] = modelview_matrix.at<double>(col,row);
}
}
i made sure the camera intrinsic matrix contains correct values, the portion which converts the opencv Mat to an opengl view matrix i believe to be correct as the cube translates and rotates in the right directions.
once the view matrix is calculated, i use it to draw the cube as follows :
_projectionMatrix = glm::perspective<float>(radians(60.0f), fabs(view.bounds.size.width / view.bounds.size.height), 0.1f, 100.0f);
_cube_ModelMatrix = glm::translate(glm::vec3(0,0,0));
const mat4 MVP = _projectionMatrix * openGLViewMatrix * _cube_ModelMatrix;
glUniformMatrix4fv(glGetUniformLocation(_cube_program, "ModelMatrix"), 1, GL_FALSE, value_ptr(MVP));
glDrawElements(GL_TRIANGLES, 36, GL_UNSIGNED_INT, BUFFER_OFFSET(0));
Is anyone able to spot my error?