Some issue about reprojection error and imu error

[Richardcvpr]

The calibtation results is not good enough. I think there may be some issues with IMU, but I don't know where the problem is.

Calibration results

Normalized Residuals

Reprojection error (cam0): mean 4.389178270980999, median 3.4851612083506827, std: 3.546836339123121 Gyroscope error (imu0): mean 4.987582349959954, median 2.840098285394952, std: 5.739188841909919 Accelerometer error (imu0): mean 5.648619958147809, median 4.5686257632248255, std: 3.9259018299971866 Residuals

Reprojection error (cam0) [px]: mean 4.389178270980999, median 3.4851612083506827, std: 3.546836339123121 Gyroscope error (imu0) [rad/s]: mean 0.04001092025534412, median 0.02278357288981692, std: 0.046040388102236 Accelerometer error (imu0) [m/s^2]: mean 0.7112548280396047, median 0.5752656676632362, std: 0.49433607707441224 Transformation (cam0):

T_ci: (imu0 to cam0): [[ 0.92757641 -0.27752758 0.25016086 -0.05011873] [ 0.30536953 0.94890312 -0.07957594 0.01329463] [-0.2152939 0.15020427 0.96492861 0.33276062] [ 0.0.0.1.]] T_ic: (cam0 to imu0): [[ 0.92757641 0.30536953 -0.2152939 0.11407051] [-0.27752758 0.94890312 0.15020427 -0.07650671] [ 0.25016086 -0.07957594 0.96492861 -0.30749456] [ 0.0.0.1.]] timeshift cam0 to imu0: [s] (t_imu = t_cam + shift) -7.978506571910605 Gravity vector in target coords: [m/s^2] [-0.1185964 0.31407217 9.80080183]Calibration configuration

cam0

Camera model: pinhole Focal length: [287.5931203196147, 318.44992107782025] Principal point: [381.33348613478404, 245.2892815651306] Distortion model: radtan Distortion coefficients: [0.014059, -0.021049, 0.001291, 0.003152] Type: aprilgrid Tags: Rows: 6 Cols: 6 Size: 0.021 [m] Spacing 0.005999999999994 [m] IMU configuration

IMU0:

Model: calibrated Update rate: 100.0 Accelerometer: Noise density: 0.01259165660479 Noise density (discrete): 0.1259165660479 Random walk: 0.000426938443672722 Gyroscope: Noise density: 0.000802210719501511 Noise density (discrete): 0.00802210719501511 Random walk: 2.82325427421011e-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]

imu_image_good_2024-06-03-17-58-28-report-imucam.pdf

Can anyone explain me, what's wrong with calibration results? I would be thankful for any answer. Thanks

[Richardcvpr]

[victorsillerico]

Some pointers: 1) You have high re-projection error, maybe the intrinsic and extrinsic parameter values are wrong. Did you use Kalibr (from ETH) for camera (only) calibration? 2) The plots of your bias clearly show that the safe bounds 3-sigma were exceed.