UK train stations trial Amazon emotion recognition on passengers

Amazon-powered AI cameras are now being used to monitor and analyze passengers’ emotions by employing a combination of “smart” CCTV cameras, which can detect objects or movements from the images they capture, along with older cameras that have their video feeds connected to cloud-based analysis software.

The initiative has been trialed in eight train stations across the UK over the last two years, with the aim of enhancing passenger safety and improving customer service. The AI cameras, integrated with Amazon’s advanced machine learning algorithms, are designed to detect age, gender and a range of emotions, WIRED reports. By interpreting facial expressions and other non-verbal cues, the system can identify passengers who may be distressed or agitated, potentially allowing officers to preempt conflicts or emergencies.

Both the social acceptance and scientific accuracy of emotion recognition are topics of ongoing controversy.

The trials are also using machine learning for object recognition to detect people trespassing on tracks, potentially predict platform overcrowding, as well as gauge antisocial behavior.

According to a network rail document released last year, trespassing is a significant problem on the railway, including at many stations. In 2022 to 2023, for the railway as a whole, the total delay directly associated with trespassing was 899,522 minutes from 21,990 incidents.

In September 2023, the London Underground completed a proof of concept at Willesden Green Tube station for an AI-assisted “Smart Station” designed to provide video analytics and real-time data insights on passenger behavior. The final report from the pilot outlines design principles for future iterations of the system and defines various use cases and triggers, such as counting customer entries and exits, and generating real-time alerts for behaviors like fare evasion, leaning over the tracks, vaping, sitting on benches for extended periods, or unfolding e-scooters.

While the partly redacted document clarifies that no facial recognition is utilized by the Smart Station platform, the pilot did involve testing additional requirements for the fare evasion scenario, including the unblurring of facial images to identify repeat offenders.

Article Topics

Amazon Web Services (AWS)  |  cctv  |  emotion recognition  |  facial analysis  |  pilot project  |  transportation  |  UK  |  video surveillance