For a regular RGB camera, reading the wikipedia page for camera resectioning/calibration here, they divide the paramters into two categories:
Intrinsic: - focal length, scale factors in x and y, and principal point
Extrinsic: - the rigid body transform of the camera relative to some origin point
My question is how does camera calibration for a depth camera relate to these parameters, for example a Kinect v2 depth camera.
Edit: For example in the SUN RGB-D dataset, there are two files associated with each png image/ png depth image pair. An intrinsic file, and an extrinsic file.
The intrinsic file is a 3 by 3 matrix, example:
529.500000 0.000000 365.000000
0.000000 529.500000 265.000000
0.000000 0.000000 1.000000
From what I understand, using 1-indexing, entry (1,1) and (2,2) according to the camera resectioning article on wikipedia are the focal length multiplied by scale factors relating "pixels to distance" (any elaboration on this appreciated). Whereas entries (1,3), (2,3) represent the coordinates of a principal point. Is this principal point in 2-D or 3D space and what do these numbers mean?
The extrinsics file/s (for some reason there are multiple files sometimes) are a 4 by 4 matrix, example:
0.979445 0.201365 0.011827 0.000000
-0.201365 0.972640 0.115861 0.000000
0.011827 -0.115861 0.993195 0.000000
Which is a rigid body transform.