How to do calibration of the ‘cam_to_pos’?
I am curious about how the cam_to_pos is determined. I think this is related to the documentation as below?
We obtain accurate georegistered vehicle poses using structure-from-motion based on all feature matches while regularizing against the GPS location and the LiDAR scans. We also release raw OXTS (GPS/IMU) measurements. Note that the poses obtained from structure-from-motion are not identical to the raw OXTS measurements.
Are ground control points used in the SFM step and then average the calibration between camera POS and GPS/IMU POS according to the same timestamp?
Hi, sorry for the late reply. To calibrate cam_to_pos, we recorded trajectories of the camera (visual odometry) and the GPS/IMU, and then solved an optimization problem. Here are some slides for more details of the calibration.