Enhancing the accuracy of rigid registration using a general and adaptive robust loss function

Yibin Xu


Supervised by Bailin Deng; Moderated by Matthew J W Morgan

Rigid registration of two geometric data sets is essential in many applications, including robot navigation, surface reconstruction and shape matching. Most commonly, variants of the Iterative Closet Point (ICP) algorithm are employed for this task. These methods alternate between closest point computations to establish correspondences between two data sets, and solving for the optimal transformation that brings these correspondences into alignment. However, a major difficulty for this approach is the sensitivity to outliers and missing data often observed in 3D scans.In this project, we consider using a general and adaptive robust loss function to enhance the Sparse Iterative Closest Point approach furthermore by allowing the algorithm to abandon some points when evaluating the losses. In this way, providing a more precise result when registering two geometric data sets.

Initial Plan (03/02/2020) [Zip Archive]

Final Report (15/05/2020) [Zip Archive]

Publication Form