A Lie algebra representation for efficient 2D shape classification

作者：

Highlights：

• This paper presents a novel solution to address high computational cost of current Riemannian framework based methods.

• We propose a novel representation of Block Diagonal Symmetric Positive Definite Matrix Lie Algebra (BDSPDMLA) for 2D shapes.

• A new classification algorithm is introduced to achieve ten thousand times increase in speed over manifold methods.

• Experimental results on five datasets demonstrate BDSPDMLA’s effectiveness in computation times and classification accuracy.

摘要

•This paper presents a novel solution to address high computational cost of current Riemannian framework based methods.•We propose a novel representation of Block Diagonal Symmetric Positive Definite Matrix Lie Algebra (BDSPDMLA) for 2D shapes.•A new classification algorithm is introduced to achieve ten thousand times increase in speed over manifold methods.•Experimental results on five datasets demonstrate BDSPDMLA’s effectiveness in computation times and classification accuracy.

论文关键词：Lie algebra,2D Shape classification,Covariance matrix,Lie group of SPD matrix

论文评审过程：Received 22 May 2021, Revised 30 August 2022, Accepted 26 September 2022, Available online 2 October 2022, Version of Record 10 October 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.109078