This video is augmented without any prior knowledge of the scene. The camera is not calibrated and there is no interaction with any human during the process.
To calculate the camera pose, first we estimate the fundamental matrix of some frames using 8 point algorithm and levenberg-marquardt optimization algorithm. Then the structure is recovered via linear triangulation and optimized using boundle adjustment.
More info at: http://jrsanchez.users.sourceforge.net/
Link to this comment:
All Comments (0)