When the laser SLAM algorithm is run, the odometer track obtained is in the radar coordinate system, and the ground_truth is under the camera coordinate system. During the trajectory evaluation, we can use the params --toCameraCoord True to transform the trajectory to the camera coordinate system.
But the external parameter matrix in the program is just to adjust the coordinate system
R_C2L = np.array([[0, 0, 1, 0],
[-1, 0, 0, 0],
[0, -1, 0, 0],
[0, 0, 0, 1]])
But KITTI's calibration file gives the transformation relationship from radar to camera coordinate system, Tr in calib.txt
Tr: 4.276802385584e-04 -9.999672484946e-01 -8.084491683471e-03 -1.198459927713e-02 -7.210626507497e-03 8.081198471645e-03 -9.999413164504e-01 -5.403984729748e-02 9.999738645903e-01 4.859485810390e-04 -7.206933692422e-03 -2.921968648686e-01
My question is, when I use this real external parameter, I get a larger error result, but using this default parameter, I can get a smaller error result. How can I evaluate the laser slam algorithm correctly?
The first evaluation uses the external parameters of the calibration file, and the second evaluation uses the default parameters in the program
When the laser SLAM algorithm is run, the odometer track obtained is in the radar coordinate system, and the ground_truth is under the camera coordinate system. During the trajectory evaluation, we can use the params --toCameraCoord True to transform the trajectory to the camera coordinate system.
But the external parameter matrix in the program is just to adjust the coordinate system
R_C2L = np.array([[0, 0, 1, 0],
[-1, 0, 0, 0],
[0, -1, 0, 0],
[0, 0, 0, 1]])
But KITTI's calibration file gives the transformation relationship from radar to camera coordinate system, Tr in calib.txt
Tr: 4.276802385584e-04 -9.999672484946e-01 -8.084491683471e-03 -1.198459927713e-02 -7.210626507497e-03 8.081198471645e-03 -9.999413164504e-01 -5.403984729748e-02 9.999738645903e-01 4.859485810390e-04 -7.206933692422e-03 -2.921968648686e-01
My question is, when I use this real external parameter, I get a larger error result, but using this default parameter, I can get a smaller error result. How can I evaluate the laser slam algorithm correctly?
The first evaluation uses the external parameters of the calibration file, and the second evaluation uses the default parameters in the program