This may be a very simple question so please bear with me...

Suppose I have a stationary object and point clouds, $n$ $(x,y,z)$ points, of that object taken by a moving camera at time steps $t_i$, $t_{i+1}$, ...

Using ICP on the point clouds from subsequent time steps I get the relative transformation (a rotation matrix and translation vector $\in SE3$) needed to map the cloud from some $t_i$ to to the cloud at $t_{i + 1}$.

Is the Camera trajectory from time $t_i$ to $t_{i + 1}$ the same? ie I simply take the position of the camera at $t_i$ and multiply it by the rotation matrix and then add the translation vector to get the new position?


Yes that is correct. Easiest way is probably to work with the homogeneous 4x4 Tranform Matrix($T$) composed of $\begin{bmatrix}R & t\\0 & 1\end{bmatrix}$. Then your new pose is then just $T_i$ multiplied by $\Delta T_{i+1}^{i}$(relative transform). For every new relative transform just do this concatenation.

So $T_{i+1}=T_i*\Delta T_{i+1}^{i}$


Depending on your ICP implementation you may actually get the inverse relative transform $\Delta T_{i}^{i+1}$. If this happens then just invert the matrix and continue as normal.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.