goldbattle
commented
1 year ago It looks like it didn't optimize probably. Maybe the IMU noises you are using could be better.
On Tue, Mar 21, 2023 at 5:03 AM EugenKarasev @.***> wrote:
Hello! I test kalibr to calibrate my monocamera (30 fps bgr ) - imu (~49 Hz) system and i got these strange results. Should accel errors and angular velocities errors be almost zero? Can time delay deviations lead to poor calibration results?
Calibration results
Normalized Residuals
Reprojection error (cam0): mean 0.390402181164542, median 0.3190253610101713, std: 0.34872396538201356 Gyroscope error (imu0): mean 8.638410884293534e-08, median 4.105800151162524e-10, std: 3.5084008002926186e-06 Accelerometer error (imu0): mean 7.797051162952767e-07, median 7.962190946218488e-11, std: 3.098689588330113e-05 Residuals
Reprojection error (cam0) [px]: mean 0.390402181164542, median 0.3190253610101713, std: 0.34872396538201356 Gyroscope error (imu0) [rad/s]: mean 1.2093775238010947e-10, median 5.748120211627534e-13, std: 4.9117611204096645e-09 Accelerometer error (imu0) [m/s^2]: mean 1.0915871628133875e-08, median 1.1147067324705883e-12, std: 4.3381654236621587e-07 Transformation (cam0):
T_ci: (imu0 to cam0): [[-0.00000163 -0.99924041 0.03896932 -0.00002999] [-0.23080183 -0.0379168 -0.97226171 0.00019168] [ 0.97300078 -0.00899578 -0.23062646 -0.00002362] [ 0. 0. 0. 1.]] T_ic: (cam0 to imu0): [[-0.00000163 -0.23080183 0.97300078 0.00006723] [-0.99924041 -0.0379168 -0.00899578 -0.00002291] [ 0.03896932 -0.97226171 -0.23062646 0.00018209] [ 0. 0. 0. 1.]] timeshift cam0 to imu0: [s] (t_imu = t_cam + shift) -0.06078444965185353 Gravity vector in target coords: [m/s^2] [-0.34105986 -9.79777889 0.23585982]Calibration configuration
cam0
Camera model: pinhole Focal length: [1039.5099570143711, 1042.2195539224472] Principal point: [618.1704778515369, 290.37109512978014] Distortion model: radtan Distortion coefficients: [-0.3935934462048003, 0.16992322720688555, 0.003903623112852698, -0.00036856266440390197] Type: aprilgrid Tags: Rows: 7 Cols: 7 Size: 0.019 [m] Spacing 0.0038 [m] IMU configuration
IMU0:
Model: calibrated Update rate: 49.0 Accelerometer: Noise density: 0.002 Noise density (discrete): 0.014 Random walk: 0.0004 Gyroscope: Noise density: 0.0002 Noise density (discrete): 0.0014000000000000002 Random walk: 1e-05T_ib (imu0 to imu0) [[1. 0. 0. 0.] [0. 1. 0. 0.] [0. 0. 1. 0.] [0. 0. 0. 1.]] time offset with respect to IMU0: 0.0 [s]
[image: accels] https://user-images.githubusercontent.com/50025549/226557729-7e350819-8a0a-4b57-b63f-6755fd98ad89.png [image: accels_bias] https://user-images.githubusercontent.com/50025549/226557744-ab990f5f-a84a-43f1-af24-36615a678b49.png [image: accels_err] https://user-images.githubusercontent.com/50025549/226557784-a76b75ae-b270-4279-be74-2fe238d09ea0.png [image: ang_vel] https://user-images.githubusercontent.com/50025549/226557815-b82b5dcc-f351-4738-8213-576d7f5c44fc.png [image: ang_vel_err] https://user-images.githubusercontent.com/50025549/226557849-7ed478db-77e2-4cbe-abdc-efad731c42bc.png [image: cam_proj_err] https://user-images.githubusercontent.com/50025549/226557866-d47e0afd-9f1d-480a-ba63-3c4a8329d44b.png [image: gyro_bias] https://user-images.githubusercontent.com/50025549/226557899-4b4ff7a8-36e1-4020-a76c-f1a2dc6d8400.png [image: mes_rates] https://user-images.githubusercontent.com/50025549/226557919-467e9689-aa2f-4aa7-9632-3f728406691a.png [image: trajectory] https://user-images.githubusercontent.com/50025549/226557945-3169240a-293b-4774-b688-fae12f71ebc3.png
— Reply to this email directly, view it on GitHub https://github.com/ethz-asl/kalibr/issues/600, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQ6TYTIKF7Z5IH3QZPXAT3W5FVENANCNFSM6AAAAAAWCEJKK4 . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hello! I test
kalibrto calibrate my rolling-shutter monocamera (30 fps bgr) - imu (~49 Hz) system and i got these strange results.