search

hengli / camodocal

CamOdoCal: Automatic Intrinsic and Extrinsic Calibration of a Rig with Multiple Generic Cameras and Odometry
Other
1.18k stars 390 forks source link

What's camera mask and extrinsic estimates in extrinsic_calib exe? #67

Open blairlpp opened 7 years ago

blairlpp commented 7 years ago

And what kind of pictures should I take for a two cameras rig?

stanathong commented 7 years ago

I know this is a late answer.

You simply take the picture of the environment. I would suggest that you make sure that the ground is flat (less or no slope). If your cameras are on a vehicle, drive it around in the way that both camera can see the same scene e.g. turn around. If possible, you might place markers or objects in the scene so that you have more salient features which are helpful for feature detection and matching.

Shine-yzhao commented 6 years ago

Excuse me. Can I calibrate a cameras rig only with the synchronous pictures and intrinsic calibration data? @blairlpp @stanathong

stanathong commented 6 years ago

You definitely need intrinsic camera parameter for all the cameras in the rig in order to do this.

What do you mean by synchronous pictures? You have to make sure that all the cameras in the rig are timely synchronised i.e. images from each camera are captured at the same time.

Shine-yzhao commented 6 years ago

You definitely need intrinsic camera parameter for all the cameras in the rig in order to do this.

What do you mean by synchronous pictures? You have to make sure that all the cameras in the rig are timely synchronised i.e. images from each camera are captured at the same time.

Thanks for your reply. Can I calibrate extrinsic parameters if I only have Intrinsic parameters and timely synchronized photos? I have Intrinsic calibration data (.yaml .dat) and timely synchronized images from each camera, but I do not have the IMU or GPS data.

stanathong commented 6 years ago

Unfortunately, CamOdoCal requires you to have GPS and IMU as its input (though orientation/IMU data can be derived approximately from GPS data i.e. heading/yaw angle)

Shine-yzhao commented 6 years ago

Unfortunately, CamOdoCal requires you to have GPS and IMU as its input (though orientation/IMU data can be derived approximately from GPS data i.e. heading/yaw angle)

Thank you, I understand.