Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Transfer Learning Approach for Railway Technical Map (RTM) Component Identification

Obadage Rochana Rumalshan, Pramuka Weerasinghe, Mohamed Shaheer, Prabhath Gunathilake, Erunika Dayaratna

TL;DR

The paper addresses digitizing Railway Technical Maps (RTMs) from PDF formats by detecting map components and associating them with mileposts to produce structured CSV data. It compares transfer-learned object detectors (SSD, YOLOv3, Faster-RCNN) and couples the best-performing detector with an OCR stage augmented by a distortion-removal preprocessing pipeline to extract text from detected regions. The approach yields high-accuracy RTM component detection (Faster-RCNN) and an automated CSV export per image, including a method to resolve milepost–component associations under layout variability. Using a dataset of 57 labeled RTM images across eight component types, the method demonstrates strong per-class performance and offers a scalable path for broader RTM digitization in railway management. Future work will expand object types and refine the associativity algorithm for more complex RTM layouts.

Abstract

The extreme popularity over the years for railway transportation urges the necessity to maintain efficient railway management systems around the globe. Even though, at present, there exist a large collection of Computer Aided Designed Railway Technical Maps (RTMs) but available only in the portable document format (PDF). Using Deep Learning and Optical Character Recognition techniques, this research work proposes a generic system to digitize the relevant map component data from a given input image and create a formatted text file per image. Out of YOLOv3, SSD and Faster-RCNN object detection models used, Faster-RCNN yields the highest mean Average Precision (mAP) and the highest F1 score values 0.68 and 0.76 respectively. Further it is proven from the results obtained that, one can improve the results with OCR when the text containing image is being sent through a sophisticated pre-processing pipeline to remove distortions.

Transfer Learning Approach for Railway Technical Map (RTM) Component Identification

TL;DR

The paper addresses digitizing Railway Technical Maps (RTMs) from PDF formats by detecting map components and associating them with mileposts to produce structured CSV data. It compares transfer-learned object detectors (SSD, YOLOv3, Faster-RCNN) and couples the best-performing detector with an OCR stage augmented by a distortion-removal preprocessing pipeline to extract text from detected regions. The approach yields high-accuracy RTM component detection (Faster-RCNN) and an automated CSV export per image, including a method to resolve milepost–component associations under layout variability. Using a dataset of 57 labeled RTM images across eight component types, the method demonstrates strong per-class performance and offers a scalable path for broader RTM digitization in railway management. Future work will expand object types and refine the associativity algorithm for more complex RTM layouts.

Abstract

The extreme popularity over the years for railway transportation urges the necessity to maintain efficient railway management systems around the globe. Even though, at present, there exist a large collection of Computer Aided Designed Railway Technical Maps (RTMs) but available only in the portable document format (PDF). Using Deep Learning and Optical Character Recognition techniques, this research work proposes a generic system to digitize the relevant map component data from a given input image and create a formatted text file per image. Out of YOLOv3, SSD and Faster-RCNN object detection models used, Faster-RCNN yields the highest mean Average Precision (mAP) and the highest F1 score values 0.68 and 0.76 respectively. Further it is proven from the results obtained that, one can improve the results with OCR when the text containing image is being sent through a sophisticated pre-processing pipeline to remove distortions.
Paper Structure (14 sections, 8 figures, 3 tables)

This paper contains 14 sections, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background and Literature
  3. Methodology
  4. Approach
  5. Dataset
  6. Image labelling
  7. Model Development
  8. Optical Character Recognition
  9. Enhancing OCR
  10. Components and text association with corresponding milepost
  11. Automated CSV generation process
  12. Results And Discussion
  13. Object Detection
  14. Conclusions

Figures (8)

  • Figure 1: (a) A Railway Technical Map (b) Selected components for labelling.
  • Figure 2: Complete System Architecture
  • Figure 3: Distortion Removal Pipeline to enhance OCR (Example)
  • Figure 4: Solving Component Associativity Problems - Flow Diagram
  • Figure 5: Solving Component Associativity Problems - Flow Diagram
  • ...and 3 more figures