Hi professors and everyone,
I've got poor Odom-cam extrinsic results running the extrinsic_calib, and the results are not stable.
In my case, I used Scaramuzza camera model, having 3 cameras(front, left, right side of the car), and I got poor results as the left and right camera positions are not symmetry(they should be...). And also, results of multiple running are not stable.
I also tried to use only the "CamOdoCalibration" class and an external VO(ORB-SLAM2) to compute the extrinsic, and I have had relatively good results this way. And I used these results to replace the intermediate stage 0 results and rerun the extrinsic_calib with "begin-stage=1"(so it reads the replaced extrinsics), but finally I still got bad results.
Any suggestions on getting better extrinsic calibration results?
Hi professors and everyone, I've got poor Odom-cam extrinsic results running the extrinsic_calib, and the results are not stable. In my case, I used Scaramuzza camera model, having 3 cameras(front, left, right side of the car), and I got poor results as the left and right camera positions are not symmetry(they should be...). And also, results of multiple running are not stable. I also tried to use only the "CamOdoCalibration" class and an external VO(ORB-SLAM2) to compute the extrinsic, and I have had relatively good results this way. And I used these results to replace the intermediate stage 0 results and rerun the extrinsic_calib with "begin-stage=1"(so it reads the replaced extrinsics), but finally I still got bad results. Any suggestions on getting better extrinsic calibration results?