Efficient and Robust Registration on the 3D Special Euclidean Group

  • 2019-04-11 03:52:42
  • Uttaran Bhattacharya, Venu Madhav Govindu
  • 25

Abstract

We present an accurate, robust and fast method for registration of 3D scans.Our motion estimation optimizes a robust cost function on the intrinsicrepresentation of rigid motions, i.e., the Special Euclidean group$\mathbb{SE}(3)$. We exploit the geometric properties of Lie groups as well asthe robustness afforded by an iteratively reweighted least squaresoptimization. We also generalize our approach to a joint multiview method thatsimultaneously solves for the registration of a set of scans. We demonstratethe efficacy of our approach by thorough experimental validation. Our approachsignificantly outperforms the state-of-the-art robust 3D registration methodbased on a line process in terms of both speed and accuracy. We also show thatthis line process method is a special case of our principled geometricsolution. Finally, we also present scenarios where global registration based onfeature correspondences fails but multiview ICP based on our robust motionestimation is successful.

 

Introduction (beta)

None

 

Conclusion (beta)

None