Deep-learning-based recognition of symbols and texts at an industrially applicable level from images of high-density piping and instrumentation diagrams

作者：

Highlights：

• This study proposes a method for deep learning-based symbol and text recognition.

• P&ID symbol taxonomy is defined by referring to the ISO 10628-2 and PIP PIC001.

• Large and small symbols are detected individually using GFL with adaptive NMS.

• Text region is detected using EasyOCR, and the text is recognized using Tesseract.

• The method can recognize different sizes and shape complexities of high-density P&IDs.

摘要

•This study proposes a method for deep learning-based symbol and text recognition.•P&ID symbol taxonomy is defined by referring to the ISO 10628-2 and PIP PIC001.•Large and small symbols are detected individually using GFL with adaptive NMS.•Text region is detected using EasyOCR, and the text is recognized using Tesseract.•The method can recognize different sizes and shape complexities of high-density P&IDs.

论文关键词：Deep learning,High density,Piping and instrumentation diagrams,Object recognition,Symbols,Texts

论文评审过程：Received 8 February 2021, Revised 17 May 2021, Accepted 2 June 2021, Available online 9 June 2021, Version of Record 11 June 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115337