Practical globally optimal consensus maximization by Branch-and-bound based on interval arithmetic

作者：

Highlights：

• We achieve globally optimal consensus maximization by Branch-and-Bound framework and draw the idea of interval arithmetic-based bound calculation back on the map.

• We provide the detailed derivation of interval arithmetic-based bound calculation for consensus maximization problems with both linear and quasi-convex residuals.

• Extensive experiments show that the proposed method can better deal with larger number of data points and higher outlier ratios than existing global methods.

摘要

•We achieve globally optimal consensus maximization by Branch-and-Bound framework and draw the idea of interval arithmetic-based bound calculation back on the map.•We provide the detailed derivation of interval arithmetic-based bound calculation for consensus maximization problems with both linear and quasi-convex residuals.•Extensive experiments show that the proposed method can better deal with larger number of data points and higher outlier ratios than existing global methods.

论文关键词：Consensus maximization,Globally optimization,Robust model fitting,Branch-and-bound,Interval arithmetic

论文评审过程：Received 17 April 2020, Revised 19 January 2021, Accepted 8 February 2021, Available online 16 February 2021, Version of Record 24 March 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107897