Skip to main content

Machine Learning Solutions

Digital Transit Limited has developed a wide variety of machine learning software for use in the transport industry.

Features include:

  • Fully autonomous video analytics system
  • Makes use of cutting-edge convolutional neural networks (CNNs) – artificial intelligence
  • Able to detect objects of interest from standard video
  • Integrates easily with customers’ existing systems
  • Works with existing cameras or new camera deployments
  • Works with off-the-shelf hardware

High-standard systems:

  • Hardware is fully approved for fitment to rolling stock
    • Shock/vibration/fire certified to EN50155
    • EMC test certification
  • Software is developed and validated to EN50657 basic integrity for rolling stock under ISO9001/IS090003 control.

Assist

Detects users with accessibility issues and either issues directions, or alerts staff.

Our object detection system utilises the cutting edge in convolutional neural networks and deep learning to autonomously spot any passengers who might be in need of assistance. This information can be simply and quickly sent to platform/train staff, who can provide assistance when and where it is needed, or to other additional systems. All processing is done station-wide in real-time across multiple simultaneous CCTV feeds.

Our object detection system utilises the cutting edge in convolutional neural networks and deep learning to autonomously detect bags and passengers, linking them together. If those bags are then left behind by a passenger, or taken by another passenger, this information can be simply and quickly sent to platform/train staff, who can provide assistance when and where it is needed, or to other additional systems. All processing is done station-wide in real-time across multiple simultaneous CCTV feeds.

FAST

Uses machine learning to detect when bags have been left behind or stolen, and sends alerts to staff.

Forward

Detects track workers and their adherence to procedures, signals and other obstructions using the forward-facing train camera.

The Intelligent Vision System is able to spot a myriad of objects that are seen by train drivers. This includes speed signs, track workers, signals, trespassers and other obstructions, all detected in real-time, using existing cameras placed on the train.

PantoView is software that uses an end-to-end neural network to determine the acceleration of the pantograph head using only a video feed. It will minimise manual operations by reducing the need for manual checks and allowing longer maintenance periodicities.

Pantoview

Measures pantograph acceleration using video footage to determine when maintenance is required.

People Counting

Provides a live count of the number of passengers per carriage on public transit.

Utilising an intelligent vision system and the CCTV cameras currently in use on public transport, the number of passengers aboard a given tram are counted. This data can be viewed in real time, allowing an operator knowledge of the total number of passengers aboard each carriage.

The Intelligent Vision System is computationally efficient in reducing data dimensionality in detecting objects on a train platform. It autonomously performs platform train interface (PTI) monitoring by detecting when a person or object is over the yellow line, helping to reduce vehicle dwell times.

Trap & Drag Detection System

Detects trap-and-drag incidents using exterior train cameras.

SEW

Repairs and stitches the lifecycle closely together

SEW is a tool that uses machine learning to highlight and quantify errors in the artefacts produced during complex product development and project delivery. SEW takes hazard logs and requirements documents as inputs, and can identify errors, low quality requirements, and other non-conformances within the documents. It outputs quality metrics so that the health of a complex project can be assessed.