A sparse structure for fast circle detection

作者：

Highlights：

• A novel formulation: The formulation tries to cover each circle instance by a pre-determined number of maximally compatible edge points.

• A time-saving decomposition: We decompose the circle detection into radius-dependent and -independent part.

• A sparse structure: We explore the statistical sparsity behind the radius-independent part and design a sparse structure for its calculation.

• A 3D voting scheme: Calculation of the radius-independent part is then implemented via a 3D voting that can be updated in a very fast manner.

• A comprehensive dataset: We construct a dataset composed of 5 categories that present considerable diversities.

摘要

•A novel formulation: The formulation tries to cover each circle instance by a pre-determined number of maximally compatible edge points.•A time-saving decomposition: We decompose the circle detection into radius-dependent and -independent part.•A sparse structure: We explore the statistical sparsity behind the radius-independent part and design a sparse structure for its calculation.•A 3D voting scheme: Calculation of the radius-independent part is then implemented via a 3D voting that can be updated in a very fast manner.•A comprehensive dataset: We construct a dataset composed of 5 categories that present considerable diversities.

论文关键词：Circle detection,Hough transform,Voting,Sparse structure,Oriented chamfer distance

论文评审过程：Received 29 December 2018, Revised 11 July 2019, Accepted 25 August 2019, Available online 26 August 2019, Version of Record 3 September 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.107022