Ask Your Question
0

opencv triangulatePoints() giving NaN values

asked 2017-09-26 01:50:56 -0600

ambikav gravatar image

I'm trying to implement standard Structure from Motion code but the triangulatePoints is resulting in NaN values. I do the following steps -

  1. get ORB features for image 1 and image 2.
  2. Match the features.
  3. get essential matrix from the matched features and camera intrinsic info available, using findEssentialMat()
  4. using recoverPose() to get pose of camera 2. Camera 1 is assumed to be at the origin.
  5. get projection matrices for camera1 and camera2.
  6. use triangulatePoints() to get 3D points.

I have tried the Matlab triangulate() function as well with the projection matrices for camera1 and camera2 as well as matched points obtained above. In case of Matlab I get good results. But using opencv in C++ (Opencv 3.3) I get NaN values.

edit retag flag offensive close merge delete

Comments

I'm not sure what's wrong with triangulate, but you can try using THIS module, which can handle more than two cameras.

Can you post your projection matrices, a few points, and the results you expect?

Tetragramm gravatar imageTetragramm ( 2017-09-26 18:30:15 -0600 )edit

Here are the two matrices - Cam1 = [1000 0 500 0 0 1000 500 0 0 0 1 0]

cam2 = [720.6221455223817, -103.3307817561171, 848.5437366007654, -951.8867514749264; -177.5149711798716, 950.5881039101392, 561.1333992122866, -148.7626862924559; -0.3953628981993094, -0.08757713995677291, 0.9143404307392484, 0.0602672098669313]

A few points - image1 - [363, 240] [226, 308] [204, 308] [217, 310] [260, 281]

image2 - [325.2, 264] [157.2, 324] [132, 324] [147, 325] [198, 301.2]

result i expect (that i get from matlab) - -0.2931 -0.55565 2.1387 -0.5184 -0.36494 1.893 -0.54948 -0.35738 1.8569 -0.53127 -0.35903 1.8787 -0.46763 -0.42688 1.9486

ambikav gravatar imageambikav ( 2017-10-02 16:46:17 -0600 )edit

1 answer

Sort by ยป oldest newest most voted
0

answered 2017-10-02 17:54:07 -0600

Tetragramm gravatar image

So I didn't have an problems, using the latest version. Only very small errors, e-5, and e-6. Are you sure you are passing the correct inputs and normalizing the outputs?

Mat_< float > Cam1 = ( Mat_< float >( 3, 4 ) << 1000 , 0 , 500 , 0 , 0 , 1000 , 500 , 0 , 0 , 0 , 1 , 0 );
Mat_< float > Cam2 = ( Mat_< float >( 3, 4 ) << 720.6221455223817 , -103.3307817561171 , 848.5437366007654 , -
                       951.8867514749264 , -177.5149711798716 , 950.5881039101392 , 561.1333992122866 , -
                       148.7626862924559 , -0.3953628981993094 , -0.08757713995677291 , 0.9143404307392484 ,
                       0.0602672098669313 );
vector< Point2f > ptsA{
    Point2f( 363, 240 ), 
    Point2f( 226, 308 ), 
    Point2f( 204, 308 ), 
    Point2f( 217, 310 ), 
    Point2f( 260, 281 )
};
vector< Point2f > ptsB{
    Point2f( 325.2, 264 ),
    Point2f( 157.2, 324 ), 
    Point2f( 132, 324 ), 
    Point2f( 147, 325 ), 
    Point2f( 198, 301.2 )
};
vector< Point3f > expected{
    Point3f( -0.2931, -0.55565, 2.1387 ), 
    Point3f( -0.5184, -0.36494, 1.893 ), 
    Point3f( -0.54948, -0.35738, 1.8569 ), 
    Point3f( -0.53127, -0.35903, 1.8787 ), 
    Point3f( -0.46763, -0.42688, 1.9486 )
};
Mat exp(expected);
exp = exp.reshape( 1, expected.size() );

Mat res;
triangulatePoints( Cam1, Cam2, ptsA, ptsB, res );
res.row( 0 ) /= res.row( 3 );
res.row( 1 ) /= res.row( 3 );
res.row( 2 ) /= res.row( 3 );
res.row( 3 ) /= res.row( 3 );
res = res.rowRange( 0, 3 ).t();
for ( int i = 0; i < res.rows; ++i )
    cout << norm( res.row( i ), exp.row( i ) ) << "\n";
edit flag offensive delete link more

Question Tools

1 follower

Stats

Asked: 2017-09-26 01:50:56 -0600

Seen: 410 times

Last updated: Oct 02 '17