Visualizing Vitality: Decoding P&ID Connectivity via Symbol Detection and Line Analysis
M R Pillai, L N Teja Tadepalli, S R Bhamre, Shreya M Gajbhiye, Deepak S Uplaonkar · 2024
The research utilizes YOLOv5 for symbol detection, EasyOCR for text recognition, and the Progressive Probabilistic Hough Transform (PPHT) for identifying straight lines in Piping and Instrumentation Diagrams (P&IDs), pivotal blueprints depicting industrial plant piping, instrumentation, and control systems. During the training phase, symbols are meticulously cropped and labeled using an open-source tool. Performance evaluation reveals that early stopping applied to the Group Shuffle Split method achieves an outstanding precision of 93%, surpassing alternative approaches. These results underscore the effectiveness of our sampling and model refinement strategies in optimizing P&ID analysis. Identified texts and symbols are cataloged based on unique IDs, occurrence frequency, and labels, enhancing comprehension crucial for efficient industrial facility operation and maintenance. Additionally, PPHT aids in detecting lines between symbols identified by the YOLOv5 algorithm, further enriching the analysis of P&IDs.