The UK Center for Ecology & Hydrology (UKCEH), working with Keen AI and Network Rail, have come up with an interesting approach to improve the environment around railway lines — by using on-board cameras and artificial intelligence technology to monitor flora and fauna.
"Our equipment was able to take thousands of clear images from a train traveling at up to 80 miles per hour," explains Tom August, PhD, a computational ecologist at UKCEH and one of the team working on the project, "and our AI software can identify ash and other species to a high level of accuracy."
"Network Rail spends £200 million (around $252 million) each year on vegetation management in order to keep the network operational," adds Amjad Karim, chief executive officer of Keen AI. "The aim of our work is to give staff at Network Rail the tools they need to safely and accurately identify where action may be required.
"We've been pushing the boundaries of what is possible when it comes to the speed of the camera, quality of images and size of the system, all while keeping it flexible and low-cost."
The system is designed to monitor biodiversity by railway tracks, automatically identifying different tree and plant species — including the invasive Japanese knotweed, which Network Rail says is responsible for a number of complaints every year, Himalayan balsam, and poisonous ragwort.
Two initial trails have proven the system capable of delivering, though work now continues on improving its performance — including increasing the rate at which images are captured, improving the precision of geolocation, and supporting train speeds of up to 100 miles per hour. The train-based camera systems are also supported by fixed-location monitoring stations that record images and audio to track animals in the area.
"The partnership with UKCEH and Keen-AI has shown that using AI can be a safer, quicker, more-cost-effective and more comprehensive way of monitoring land surrounding the railway," claims Neil Strong, PhD, Network Rail's manager for biodiversity strategy, "and we’re excited to see how this technology can be developed further to help us realize our ultimate goal of achieving a biodiversity net gain by 2035."
The team has not yet publicly published its work.