Saturday, 14 June 2014

Estimate Homography to do 2D to 3D integration (26): Successfully accessing EPnP algorithm based on Kinect Camera only between 3D and 2D points

I. Summary of preliminary results:

    1. Mapping results:
       (A) 'GT': ground truth points
             'E': coordinate of points by using EPnP alg. without optimization step
             'EGN': coordinate of points by using EPnP alg. with 'Gaussian_Newton' optimization step
         
             '+': coordinates in World Coordinate System
             '*': coordinates in Projection Coordinate System


    2. Information of coordinates:




    3. Information of the transformation matrices in Rotation and Translation:

Information related to this transformation without optimization process

Information related to this transformation with optimization process

II. Comparison of results based on previous and the current Intrinsic Parameters:

    1. Comparing Intrinsic Parameters:

Previous (...directly using built-in MATLAB calibration toolbox)

Current  (... using the MATLAB calibration toolbox proposed by Caltech/bouguetj)

    2. Transformation results provided by EPnP alg.:

        (A) 'GT': ground truth points
              'E': coordinate of points by using EPnP alg. without optimization step
              'EGN': coordinate of points by using EPnP alg. with 'Gaussian_Newton' optimization step
         
             '+': coordinates in World Coordinate System

Previous result based on built-in MATLAB calibration toolbox

Current result based on the MATLAB calibration toolbox proposed by Caltech/bouguetj



No comments:

Post a Comment