Download PDFOpen PDF in browser

Spotting Railway Signs to Build Smart Decision Support Tools in Railway Management Systems

EasyChair Preprint no. 8456

5 pagesDate: July 12, 2022


Railway being an important mode of transportation, it demands highly precise management and decision support as it is extensively used for both commuter and cargo transportation. Also, it is considered as a salient element in smart city and modern infrastructure planning. Railway track diagrams are available with service providers in portable document format where a single document consists of information from one station to another, including information regarding the tracks, signals, crossovers, switches, and their location details denoted using a standard set of symbols and drawn using a computer aided tool. A Management tool that has all details of individual symbols is an important tool for decision support systems. This research focuses on developing an automated system to extract this information based on deep learning techniques. The method consists of two steps: object detection and optical character recognition (OCR). State of the art Convolutional Neural Network (CNN) architectures are used to perform object detection. They include single-stage detectors like YOLOv3 and SSD and two-stage detectors like Faster-RCNN and RFCN. Among the selected, RFCN resulted in the highest accuracy with the minimum loss value of 0.22, compared to other methods. This is because of RFCNs architecture catering small object detection by dividing the image into small feature maps. Then, OCR is performed on detected Regions of Interest (RoI) to extract and store the text in a dedicated database which has the information of all the signs along with their location details. Image processing techniques such as template matching and Neural Network (NN) based OCR is tested here. Out of these two approaches, NN based technique outperformed template matching drastically with more than 50% accuracy.

Keyphrases: decision support tools, object detection, Railway Management Systems, railway sign detection

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
  author = {Pramuka Weerasinghe and Mohamed Shaheer and Rochana Rumalshan and Prabhath Gunathilake and Erunika Dayaratna},
  title = {Spotting Railway Signs to Build Smart Decision Support Tools in Railway Management Systems},
  howpublished = {EasyChair Preprint no. 8456},

  year = {EasyChair, 2022}}
Download PDFOpen PDF in browser