P&ID Recognition and SOP Cross Referencing

This project provides an automated pipeline for processing Piping & Instrumentation Diagrams (P&IDs) PDFs. It converts PDFs into high-resolution images, extracts text boxes using OCR, compares extracted numerical values (e.g., PSIG and Temperature) against standard limits defined in a Standard Operating Procedure (SOP) document, and then annotates the images to highlight discrepancies.

---

Features

• OCR-Based Text Extraction: Uses Tesseract OCR to identify and extract text boxes from images.
• Discrepancy Checking: Compares extracted numerical values against predefined limits.
• Annotated Results: Draws color-coded rectangles on images — green for values within limits and red for values exceeding limits.

---

Prerequisites

• Python 3.12
• Tesseract-OCR: Download and install Tesseract OCR. Make sure it’s in your system PATH or configure the path in your code.
• Poppler: Required for PDF-to-image conversion.
  • Ubuntu: sudo apt-get install poppler-utils
  • Windows/Mac: Download Poppler and add to your PATH.

Python Libraries

Install the required Python packages via pip:

pip install opencv-python pytesseract pdf2image python-docx

---

Directory Structure

diagram-analysis/
├── data/
│   ├── p&id/           # Directory for PDF diagram files
│   └── sop/            # Directory for SOP DOCX files
├── check_discrepancy.py
├── config.py
├── extract_text_box.py
├── load_sop.py
├── main.py
├── pdf_to_image.py
└── README.md

---

Usage

1. Prepare Input Files:

   • Place your PDF diagram file(s) in the data/p&id/ directory.
   • Place your SOP DOCX file(s) in the data/sop/ directory.

2. Run the Program:

   Execute the main script by running:

   python3 main.py "./data/p&id/diagram.pdf" "./data/sop/sop.docx"

3. Output:

   • Annotated images are saved in an output directory named after the PDF file name.
   • The console will also display the extracted limits and the comparison results for each processed image.

---

Example Output

Annotated Diagram Result

OCR Text Extraction Result

---

Customization

• OCR Settings:
  Modify first_config and second_config in config.py to adjust Tesseract OCR performance based on your diagram quality.

• Merging Parameters:
  Change parameters such as MIN_BLOCK_DISTANCE and BOX_PADDING in config.py to fine-tune how text boxes are merged.

• Image Scaling:
  Change SCALE_FACTOR parameters in config.py to cater for different image resolutions.

---

Possible Improvements

• Better Layout Analysis: Locating component locations by grouping Tesseract's text boxes can be improved by training a YOLO model.

• Graph Construction: Thoughts are to connect modules based on horizontal/vertical alignment if can achieve higher component recognition quality.

• Limit Mapping: Mapping logic will be improved if can achieve higher component recognition quality.

---