Abstract

This paper proposes a robust registration and alignment framework for multiple point clouds using low rank and sparse decomposition. A coarse registration phase utilizing Huber-ICP is firstly performed to roughly align all the point clouds to a same location, and then sparse and low rank decomposition is applied to extract the low rank subspace of all the point clouds, which is expected to be outlier and loss data free. Finally, a fine registration procedure can be carried out between each point clouds from this low rank space to not only a more accurate registration result but also a more precise correspondence. Robustness of our method for outliers contained in point clouds is verified through manufactured data and it also shows that an effective result can still be achieved even when some points in the cloud are lost.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call