AG3line: Active grouping and geometry-gradient combined validation for fast line segment extraction

作者：

Highlights：

• Active grouping in the anchor map with line geometry constraint enables reducing the fracture caused by unstable pixels.

• Exploiting the alignment of gradient magnitude, but not the gradient orientation in previous methods, is also effective in controlling false positives.

• The state of the art methods performed differently in the two sets of ground truth that are different in line segments’ length.

• The quantitative evaluation showed that proposed method outperformed the classical line segment detectors in both the accuracy and the computation time.

摘要

•Active grouping in the anchor map with line geometry constraint enables reducing the fracture caused by unstable pixels.•Exploiting the alignment of gradient magnitude, but not the gradient orientation in previous methods, is also effective in controlling false positives.•The state of the art methods performed differently in the two sets of ground truth that are different in line segments’ length.•The quantitative evaluation showed that proposed method outperformed the classical line segment detectors in both the accuracy and the computation time.

论文关键词：Line extraction,Active grouping,Anchor map,Alignment of gradient magnitude

论文评审过程：Received 5 June 2019, Revised 17 November 2019, Accepted 16 January 2021, Available online 20 January 2021, Version of Record 28 January 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107834