Case Study

Image Analysis to Enhance Ticket Processing Software

Key Details

Challenge AI solution to automate ticket recognition and delays repay claims processing
Solution OCR-based model to enable recognition, classification, and analysis of different types of tickets
Technologies and tools Open CV, TensorFlow, Google Vision API

Client

The Client is a large train operator seeking to keep up with the tendencies and implementing AI solutions for processes automation. The company is interested in replacing paper tickets with smartcards to automate and improve processes. One of the KPIs that the Client has to perform against is effective claims management in case of train cancellation or delay. The staff has to process many delays repay claims, and deal with massive amounts of paper tickets, which requires extra expenses.

Many passengers are inert in adopting tech-driven changes. The staff still has to process many delays repay claims manually despite the option of introducing smartcards to improve routine processes. But as passengers continue using paper tickets, it becomes more complex for staff to complete the reconciliation of cash payments.

Challenge: AI ticket processing system to automate ticket recognition and claims processing

InData Labs team was challenged to provide a solution to the existing problem. The way out was the development of a fully-functional AI-driven ticket processing system. That application was aimed to reduce the administrative burden by automating the compensation claims processing. Our team provided expert guidelines on the phases of developing an application for mobile platforms that would be able to automatically recognize, classify, and validate a wide variety of train tickets.

Solution: OCR-based model to enable recognition, classification, and analysis of different types of tickets

The mobile application was expected to encompass the following modules:

  1. Image capture module. Allows acquiring images from a smartphone camera
  2. Image processing module. Enables image enhancement and normalization to make the image acceptable for OCR by utilizing the OpenCV library:
    • Improve lighting by applying background correction algorithms
    • Enable image alignment
    • Find regions of interest (ROI) and extract regions with text.
  3. OCR module. By utilizing Google Vision API, it becomes possible to detect text within an image and automatically identify the language (only supports the English language)
  4. Ticket classification
  5. Analysis of OCR output and finding relevant information
  6. Neural Network API module to get classifications and confidence scores
  7. Displaying results

The diagram below demonstrates an example of a high-level architecture designed by the InData Labs’ team:

A would-be ticket processing software could be delivered in a Docker container that included the model and the API. Another option was to implement queue to interact with the model on the API.

Result: AI consulting services and developing a project roadmap

InData Labs provided the Client with computer vision consulting services and elaborated the project roadmap to develop and deliver a first-rate solution in compliance with the Client’s specific business needs.

A would-be solution was expected to help the Client respond to various customer complaints more effectively. The solution could also become a valuable tool to optimize costs, augment repetitive tasks with AI, improve customer satisfaction, and support compliance with multiple regulations.

Autre Articles