Railways plans AI cameras; new orders on hold

NEW DELHI: Indian Railways is likely to mandate procurement of artificial intelligence-based cameras in all future train surveillance systems, with fresh orders on hold until revised guidelines are finalised.

Officials said existing supply orders will continue, but new procurements have been paused pending a review of specifications.

"AI has developed rapidly since the existing specifications for surveillance systems within trains were finalised," a senior official told ET, adding that these need a critical review.


Research Designs and Standards Organisation first specified internet protocol-based video surveillance systems for rolling stock in March 2018. The norms have since been amended seven times, most recently in June 2025.

The current specifications define parameters for closed-circuit television cameras that record activities inside and outside trains for post-event analysis.

However, these systems lack object or defect identification and do not support maintenance tracking. Their primary purpose is reactive-helping reconstruct events after an incident rather than enabling real-time prediction or prevention.

Also Read: Boarding moving trains may soon cost you Rs 2,000, RPF mulling on-the-spot fines: Report

"Now protocols will bring in advanced surveillance capabilities and some level of real-time predictability on all cameras deployed on trains," the official said.

In February, the national transporter said its AI-powered facial recognition system had identified habitual offenders at Visakhapatnam railway station, calling it a milestone in real-time identification and apprehension of offenders.

Also Read: Constructing 15 kilometres bullet train track every month: Ashwini Vaishnaw

Kapil Bardeja, co-founder and CEO of Vehant Technologies, said existing cameras can be upgraded if they meet minimum specifications on resolution, frame rate and night performance.

"The AI processing and analytics can be handled on backend servers with adequate compute power, minimising the need for complete hardware replacement," he said, adding that AI-enabled systems typically cost three to four times more than conventional surveillance systems.