A railway-specific solution based on Auto Intellect

Auto Intellect automatically recognizes numbers on cargo and passenger rail cars, detects travel direction, and counts cars in a live video stream. Resulting metadata is synced with video footage and written into the database. For each rail car, the system can create a report containing the time of checkpoint crossing, name of checkpoint, and the relevant still frames. Thanks to this, rail movements, rolling stock, and cargo are well protected whether on the move or at a rail depot.

Auto Intellect makes a difference in monitoring a railway

Auto Intellect offers a broad toolkit for automatic registration of trains and single rail cars. It logs all movements with granularity to the level of individual rail cars passing through checkpoints. You can rely on its powerful analytics to verify, search, and analyze the data:

  • automatic video recognition of numbers on cargo and passenger rail cars with a recognition accuracy estimate;
  • for each rail car, the following are recorded into the database: car number, date and time of the checkpoint crossing, travel direction, name of the checkpoint, and recognition quality estimate;
  • side and top view synchronized video recording of passing trains and their content;
  • automatic tank car level measurement using thermal cameras; level data is recorded into a database and captioned on video;
  • checking and editing recognized numbers against actual video footage, adding comments to car numbers;
  • database searching for cars, based on checkpoints, time interval, inventory number (full or partial), text comments, and tank car level;
  • generating reports per rail car with an option to attach video frames.


Key challenges in rail car numbers recognition

Poorly visible, stenciled, and inconsistently written numbers are a common occurrence. Based on the real-world experience with number recognition and in-house testing (with a sample of approximately 10,000 rail cars), AxxonSoft experts selected recognition algorithms with exceptionally high-quality results. Unlike most of the algorithms currently available on the market, the algorithms in Auto Intellect recognize even numbers that have been painted on the rail car chassis and eight-digit internal reference numbers at rail depots.

The advantages of Auto Intellect rail car number recognition technology

  • To enable flawless use of fisheye lenses, number recognition cameras are installed at a minimum offset from rail cars (~1 to 1.5 meters).
  • For maximum recognition quality, cameras can be installed on both sides of a railway checkpoint. In that case, even if one of the sides of the rail car has an illegible number, the other side may still provide a usable image. Up to 4 cameras per track can be installed (at 2 different levels).
  • The final recognition result passes through an intelligent quality check to make sure it contains a proper rail car number.
  • A beam sensor is used to separate rail cars from each other, count their number, and increase overall recognition quality.
  • Recognition of numbers on passenger rail cars is useful in various situations, such as during maintenance in depots.
  • Recognition of standard UIC rail car numbers (including flat and tank cars) as well as single line 8-digit numbers on car bodies and chassis.
  • Recognition of low contrast, stenciled, and inconsistently written car numbers.