I planned to use the opencv triangulatePoints () method returns the three-dimensional reconstruction,

1, I first calibration tool box of matlab is utilized to extract out of the same time different camera two images of the same objects are the characteristics of the matching point

2, then I will enter into the matching point characteristic of the two pictures cv2. Essential matrix findEssentialMat approach

3, I reuse decomposeEssentialmat () method to get the camera's rotation and translation matrix R1 R2 t

4, I use camera internal parameters combined with R1 R2 implementation know t calculation in addition to the camera projection matrix

5, 56 feature points and the projection matrix in the end I will enter into triangulatePoints () method returns the 56 three-dimensional coordinates of the same

I try to write your own a camera inside, input matrix and 56 feature points, a return to 56 reconstruction of three coordinates are different, it may say maybe my camera projection matrix calculation error, but I think I am the steps above no problem, please be encountered this kind of situation, if I need another method to calculate the camera projection matrix, if so, what should do, thank you!