Closed nikisim closed 1 year ago
It looks like there are some issues around 25-75 seconds in. Maybe you could process the bag just after these? The IMU also seems to have some timing problems of either dropped messages or just irregular publishing (those 10ms spikes). Your reprojection has very large errors and are non-gaussian suggesting something wrong with the selected model, camera, or dataset. One thought is if it is a webcam, it might be a rolling shutter camera which will cause issues with distortion of the collected image (Kalibr IMU-camera calibration assumes global shutter cameras).
Thank you very much for your answer!
1) Problems between 25-75 seconds. I kinda don't understand how should "Comparsion of predicted and measured specific force" plot looks like. I get how to validate other plots: if blue lines inside red bounds - it's okay :) But these est. bodyspline - I don't know how to interpretate
2) Yeah, I know that IMU hz plot should be just one line without any spread and other lines. But how can I achive this? As I understood, I need to rectify the timestamp problems, but how exactly - I have no idea. Although with IMU PixHawk this plot looks much better than IMU from A1 Unitree robot... You can see the last one below. (both IMU was calibrated with allan-variance-ros)
Sorry, if I ask some dumb questions...
One more question: is it possible to use Kalibr for imu->cam calibration with IMU 10 Hz? Or it is not going to work in this case?
A 10Hz IMU probably will not work. You should use a higher frequency IMU. You would need to look into th specific IMU driver you are using to see how to fix the timestamp problems. The posted IMU rate for the A1 looks very very bad so you should definitely take the time to look at that driver.
Hi!
I'm not sure that my calibtation results is good enough. Rotation from imu to camera seems pretty well for me, and translation between imu and cam is about 19 cm (in real life it's about 21-22 cm).
I guess, accelerometer bias and gyro bias plots are wrong. Also I'm not sure about reprojection errors of camera...
I'm using PixHawk 2.6.8 and simple Logitech web-camera (I'm just testing the calibration with that configuration)
Can anyone explain me, what's wrong with calibration results? I would be thankful for any answer.
Thanks