Understanding of KITTI dataset Odometry coordinate system transformation

Tr in the original data represents the transformation from the Velodyne coordinate system to the P0 camera

Tr=A^{P_0}_{veil}

The GroundTruth label represents the transformation relationship under the P0 coordinates

T_{diff}=A_{p_0,t_1}^{p_0,t_2}

The transformation of t0->t1 in the velodyne coordinate system is required, then:

T_{velo,t_0}^{velo,t_1} = {Tr}^{-1} \cdot T_{diff} \cdot Tr                (1)

Equivalent to

A_{P_0}^{veil} \cdot A^{p_0,t_1}_{p_0,t_0} \cdot A^{P_0}_{veil} = A_{P_0}^{veil} \cdot A^{P_0, t_1}_{w} \cdot A_{P_0,t_1}^{w} \cdot A^{P_0}_{veil} =(A_{veil}^{P_0})^{-1} \cdot(A_{ P_0,t_1}^{w}) ^{-1} \cdot A_{P_0,t_1}^{w} \cdot A^{P_0}_{veil}

Finally equivalent to

 (A_{velo,t_1}^{w})^{-1} \cdot A_{velo,t_0}^{w}                            (2)

In short, the above two transformations (1) (2) are equivalent

(1) Computational efficiency is higher

(2) The calculation is more complicated and needs to be converted to the P0 coordinate system first and then inverse

Guess you like

Origin blog.csdn.net/li4692625/article/details/131360512