Converting Math functions from Python to c++

Question

I am attempting to convert some opencv code from python to c++, and am a little lost. The python is:

if 0 < R[1,1] < 1:
    # If it gets here, the pose is flipped.

    # Flip the axes. E.g., Y axis becomes [-y0, -y1, y2].
    R *= np.array([
        [ 1, -1,  1],
        [ 1, -1,  1],
        [-1,  1, -1],
    ])

    # Fixup: rotate along the plane spanned by camera's forward (Z) axis and vector to marker's position
    forward = np.array([0, 0, 1])
    tnorm = T / np.linalg.norm(T)
    axis = np.cross(tnorm, forward)
    angle = -2*math.acos(tnorm @ forward)
    R = cv2.Rodrigues(angle * axis)[0] @ R

So far i have:

cv::Mat R;
if (0 < R.at<double>(1, 1) < 1) {

            // Flip the axes.E.g., Y axis becomes[-y0, -y1, y2].
            float mult[9] = { 1, -1, 1, 1, -1, 1,-1, 1, -1 };

            cv::Mat FlipAxes = cv::Mat(3, 3, CV_32F, mult);

            R *= FlipAxes;

                //# Fixup: rotate along the plane spanned by camera's forward (Z) axis and vector to marker's position
            cv::Vec3d forward(0, 0, 1);
            double tnorm = tvecBest / np.linalg.norm(T)
            axis = np.cross(tnorm, forward)
            angle = -2 * math.acos(tnorm @ forward)
            R = cv2.Rodrigues(angle * axis)[0] * R


        }

I am lost on these lines:

double tnorm = tvecBest / np.linalg.norm(T)
axis = np.cross(tnorm, forward)
angle = -2 * math.acos(tnorm @ forward)

what is the equivalent of np.linalg.norm in c++ opencv?

There isn't a direct one you can use cv::sqrt() and cv::pow() — Martin Beckett
– Martin Beckett, Commented Aug 22, 2018 at 15:15
Thanks, can I do this?: cv::Vec3d tvecNorm(tvecBest); cv::normalize(tvecNorm, tvecNorm); — anti
– anti, Commented Aug 22, 2018 at 15:27

PilouPili · Accepted Answer · 2018-08-22 15:32:06Z

3

the np.linalg.norm(T) is just the L2 norm :

sqrt(T[0]*T[0]+T[1]*T[1]+T[2]*T[2])

np.cross(tnorm, forward) is the cross product:

axis[0] = tnorm[1]*forward[2]-tnorm[2]*forward[1]
axis[1] = -tnorm[0]*forward[2]+tnorm[2]*forward[0]
axis[2] = tnorm[0]*forward[1]-tnorm[1]*forward[0]

answered Aug 22, 2018 at 15:32

PilouPili

2,7192 gold badges20 silver badges31 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

anti Over a year ago

Thank you! So that gives me:

cv::Vec3d forward(0, 0, 1); 			cv::Vec3d T(tvecBest); 			cv::Vec3d tnorm = sqrt(T[0] * T[0] + T[1] * T[1] + T[2] * T[2]); 			cv::Vec3d axis; 			axis[0] = tnorm[1] * forward[2] - tnorm[2] * forward[1]; 			axis[1] = -tnorm[0] * forward[2] + tnorm[2] * forward[0]; 			axis[2] = tnorm[0] * forward[1] - tnorm[1] * forward[0];

anti Over a year ago

What is the c++ for : cv::Vec3d angle = -2 * math.acos(tnorm @ forward) ?

PilouPili Over a year ago

the @ is not a mathematical operator in python. Was it / operator ?

PilouPili Over a year ago

from what I read they meant @ to be a dot product : tnorm @ forward -> tnorm[0]*forward [0]+ tnorm[1]*forward [1]+ tnorm[2]*forward [2]

PilouPili Over a year ago

and angle must be a double : double angle = -2 * acos( tnorm[0]*forward [0]+ tnorm[1]*forward [1]+ tnorm[2]*forward [2])

|

Collectives™ on Stack Overflow

Converting Math functions from Python to c++

1 Answer 1

7 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related