Continuous Scanning of the ARTC Network
The first successful delivery of automated data processing using Artificial Intelligence (AI) for strategic data capture across an entire network was provided to the Australia Rail and Track Corporation (ARTC) in January 2020 and is on-going, with an initial 3.5-year service provision. The challenge was to automate network-wide infringements and track clearance surveys.
ARTC’s Rail system
ARTC is Australia’s largest Railway network covering over 5,000 miles of network across New South Wales, South Australia, Victoria and Western Australia. The railway primarily forms an arterial backbone of Australia’s transportation network, carrying freight, grain, passenger and coal. ARTC looked to improve the current inspection regime as part of their transit space infringement management process, by adding an automated LiDAR and imagery-based system onboard the AK Car (geometry measurement train).
ARTC’s Requirements
ARTC focused their requirements on streamlining and automating the infringement and track clearance surveys across the network. Infringement surveys for any network are expensive and require intrusive access by rail workers to the rail corridor to perform manual measurements, which assist in defining the level of infringement. ARTC sought to deploy a safer and more technology-driven automated inspection solution across its national railway network.
Cordel’s End-to-end Capture and AI Processing Solution
The solution provides remote LiDAR capture capability from onboard the AK Car, with automated data processing, generating specific data insight use cases. The solution includes fit-and-forget hardware, with a permanently installed, remotely managed Mobile Laser System (MLS) including imagery as well as the onboard networking, data storage and 4G communications. The MLS hardware was installed permanently on the AK car in late January 2020 during a 2-day maintenance window, and the program successfully entered into a 2-month commissioning phase.
On successful completion of this initial phase, ARTC engineers completed the appropriate internal engineering acceptance, allowing ARTC to run an automated data capture, processing and output of track clearance and infringement survey data to update the National Infringement Register continuously. To date, over 50,000 miles of LiDAR point cloud survey data have been automatically processed, identifying infringements and track clearances across the entire 5,000-mile network.
Above image:-The core Cordel solution engine: Railhead Detection (automatically colourised in magenta), which is the foundation for multiple spin-off solutions, using a single LiDAR scan from a run on the measured track (in red), and identifying the other three tracks (in orange) and the structure (in green)
