Skip to content

ramugopal92/engineering-drawing-change-detection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Engineering Drawing Change Detection

AI-assisted engineering drawing change detection using alignment, ROI comparison, OCR extraction, and token analysis.

Overview

Engineering teams often spend significant time manually comparing drawing revisions to identify changes in views, notes, labels, and other document elements. This project presents a prototype system that automates part of that process using computer vision and OCR-based comparison techniques.

The system was developed as an experimental workflow to support engineering drawing review and change detection.

Key Features

  • Drawing alignment before comparison
  • Global pixel-based difference analysis
  • ROI-based localized comparison
  • OCR extraction from notes block
  • Token normalization and comparison
  • Change classification
  • Evaluation using precision, recall, accuracy, and confusion matrix

Core Techniques

  1. Image alignment
  2. Pixel difference detection
  3. ROI template matching
  4. OCR text extraction
  5. Engineering token comparison
  6. Similarity scoring
  7. PASS / CHANGE classification

Workflow

  1. Load old and new drawing files
  2. Align drawings
  3. Apply ROI templates
  4. Measure ROI-wise pixel change
  5. Run OCR on selected text regions
  6. Normalize extracted text
  7. Compare tokens and labels
  8. Generate final engineering change summary
  9. Evaluate performance using metrics

Example Output Areas

  • Section view change detection
  • Front face / side view comparison
  • Notes block OCR difference
  • ROI-based classification report
  • Confusion matrix and evaluation charts

Evaluation Metrics

The prototype was evaluated using:

  • Precision
  • Recall
  • Accuracy
  • ROI classification accuracy
  • Confusion matrix analysis
  • Token detection performance

Limitations

  • OCR performance depends on drawing clarity
  • Small symbol-level changes may require stronger detection logic
  • ROI templates are currently predefined
  • Performance may vary with scan quality and layout variation

Future Improvements

  • automatic dimension detection
  • weld symbol detection
  • dynamic ROI generation
  • better engineering OCR pipeline
  • change severity classification
  • integration with CAD/PDM workflows

Author

Ramu Gopal
Founder, The Tech Thinker
Senior Mechanical Design Engineer | CAD Automation | Engineering AI Workflows

Related Articles

Disclaimer

This repository is a research and portfolio prototype intended for educational and technical demonstration purposes.