Hikvision Brings Large-Scale AI to Traffic Cameras: 75% Fewer False Alarms

Hikvision has brought large-scale AI to traffic cameras, with new checkpoint cameras that spot drivers not wearing seatbelts or using a phone with 75% fewer false alarms, and incident detection cameras and servers that cut false detections by 60%. Announced on 30 July 2025, the range runs on traffic AI models built on Hikvision's Guanlan platform. Here's how the technology works and what it means for road safety and vehicle monitoring.

What's new

• Checkpoint cameras (iDS-TCV9(C)00-HI and iDS-TCV9(C)07-HIR) for compliance monitoring, such as seatbelt and mobile phone use.
• Traffic incident detection cameras (iDS-TCS802-HS, iDS-TCS907-HIR, iDS-TCS917-HI and iDS-TCS900-HI) for events such as fallen objects, pedestrians on the carriageway and stopped vehicles.
• Traffic AI servers (iDS-TSS500-H/16 and iDS-TSS300-H/04) that bring the same AI to existing road cameras, with no need to replace them.

How large-scale traffic AI works

Older traffic analytics looked for specific shapes: a phone, a seatbelt strap, a stationary car. That works in good conditions but falls apart in rain, at roadworks, or when a seatbelt is the same colour as the driver's jumper.

Hikvision's traffic models use a Transformer architecture with self-attention, the same family of AI behind modern large language models. Instead of judging one detail in isolation, the model looks at the whole scene and how its parts relate. Hikvision says the models were trained on millions of real-world situations, including wet roads, construction zones and extreme weather.

• Seatbelt detection looks at the driver's posture and any visible part of the belt, even when partly hidden, cutting false alarms by 75%.
• Phone-use detection considers hand position, the area near the ear and the driver's line of sight, not just the shape of a phone, again cutting false detections by 75%.
• Incident detection understands context, for example recognising hi-vis vests or emergency lights to classify what's happening, with 60% fewer false alarms.

Why fewer false alarms matter

Every false alarm costs an operator time to check, and too many lead to real alerts being ignored. Cutting false alarms by 60–75% means control rooms can trust the alerts they get, respond faster to genuine incidents such as debris or a broken-down vehicle in a live lane, and spend less time clearing nuisance events.

Who it helps

• Road operators and local authorities monitoring motorways, tunnels and busy junctions.
• Road safety partnerships looking at seatbelt and phone-use compliance.
• Operators with existing camera networks, who can add AI through the servers without new roadside hardware.

Vehicle recognition for businesses

These checkpoint and incident cameras are specialist traffic products. For businesses that need to recognise vehicles on their own sites, such as car parks, yards, gated developments and depots, the equivalent tool is an ANPR camera that reads number plates and can open barriers or log visits. Hikvision ANPR cameras we stock include:

• iDS-2CD7A26G0/P-IZHSY 2MP DeepinView ANPR bullet (8–32mm)
• iDS-2CD7A45G0/P-IZHSY 4MP DeepinView ANPR bullet (4.7–118mm) for long-distance capture
• iDS-2CD7A46G2/LM-IZHSY 4MP ANPR and AI platform bullet

Browse the full ANPR camera range. Plate capture depends on mounting position, lane layout and vehicle speed, so talk to us before ordering.

Our take

Traffic is one of the hardest environments for video analytics: fast movement, changing weather and lots of visual clutter. Large-scale AI that reads the whole scene rather than hunting for shapes is a real step forward, and the server option is a sensible way to modernise existing camera networks. The same type of AI is now reaching everyday products such as DeepinViewX cameras and AcuSeek NVRs.

Questions about ANPR or vehicle recognition for your site? Call 01902 902251. We ship nationally to England, Scotland and Wales.

Quick questions

Can AI cameras detect drivers using mobile phones?
Yes. Hikvision's checkpoint cameras analyse hand position, the area near the ear and line of sight, and Hikvision says this cuts false phone-use detections by 75%.

Do I need new cameras to use traffic AI?
Not necessarily. Hikvision's traffic AI servers can analyse video from existing cameras.

What is a Transformer AI model?
A type of AI that looks at all parts of an image at once and how they relate, rather than one detail at a time, which helps it understand context.

Do you sell the traffic checkpoint cameras?
They are specialist products. Contact us about traffic projects; for site ANPR, see our ANPR camera range.

Source: Hikvision UK newsroom, 30 July 2025. Figures are Hikvision's own.