Camera Extrinsics problem

[zzg-zzg]

I accidentally observed the T vectors of all the extrinsic matrices and noticed a strange phenomenon: all the T vectors are almost identical:
[5.8154437e-18 8.4172785e-01 2.1582971e+00]
[-1.1441708e-16 8.4172785e-01 2.1582971e+00]
[-7.4898319e-17 8.4172785e-01 2.1582971e+00]
[1.1770528e-17 8.4172785e-01 2.1582971e+00]
[2.7667727e-17 8.4172785e-01 2.1582971e+00]
[-5.2949489e-17 8.4172785e-01 2.1582971e+00]
[-7.9030384e-17 8.4172785e-01 2.1582971e+00]
[1.6320204e-17 8.4172785e-01 2.1582971e+00]
[3.3593800e-17 8.4172785e-01 2.1582971e+00]
[-9.4651181e-17 8.4172785e-01 2.1582971e+00]

This is puzzling to me. Why do the extrinsic matrices of a multi-camera system (before stereo rectification) have the same translation vector?

[ShunyuanZheng]

It is actually the case. The extrinsic is in w2c format, where the T vector means the position of the stage's center point in camera coordinate. You can convert the extrinsic to c2w format to view the T vector, which means the camera poses in world coordinate then.

[zzg-zzg]

Thank you for your reply, you're right. I misunderstood it as constructing the cam2world vector for the pose.