Hikvision iDS-7732NXI-M4/X 32-Channel DeepinMind Network Video Recorder, 8K HDMI

What this Hikvision DeepinMind recorder is for

A network video recorder (NVR) stores video sent by IP cameras over a network, rather than encoding an analogue signal itself. DeepinMind is Hikvision's name for this M Series of recorders with analytics built in.

The point of this model is analysis in the recorder itself, not only in the camera.

How the analytics work

• Two engines: each engine runs one intelligent algorithm, and the mode is adjustable. The channel figures are shared between them, so choose what each engine does at set-up.
• Deep learning: software trained on examples learns what a person or vehicle looks like. This cuts false alarms from tree branches, leaves, shadow, light and small animals.
• Video structuralization: the recorder pulls face, human body and vehicle features out of live video. You can then search by those details instead of scrubbing through hours of footage.
• H.265+: a video compression method that needs less storage than older formats for a similar picture.

Where to use it

This class of recorder usually sits in a comms room or cabinet, with a monitor where staff can see it. The working range is -10 to 55° C, so keep it out of unheated roof spaces.

Face data counts as personal data, so take advice before you turn on facial recognition. The guide to CCTV rules in the UK is a good starting point.

What the figures mean in practice

• 32-ch and 32 MP: these are separate limits. With ultra HD resolution mode enabled, the highest 32 MP/24 MP streams are limited to 8 channels.
• 320 Mbps incoming: add up the bitrate of every camera, and stay within this figure.
• 2 MP analytics figures: the 24-ch and 16-ch analytics limits are at 2 MP, with up to 8 MP beside them. Check how many channels run at 8 MP at set-up.
• Decoding capability: this sets how many streams can be shown or played at once, not how many can be recorded. It runs from 2-ch@32 MP to 32-ch@1080p.
• Face library: the list library holds up to 100,000 face pictures, and comparison from cameras runs at 48 pictures per second.

Installing it

• Cameras: cameras reach the NVR through a switch. Plan a PoE switch or another power source for the cameras, and size its power budget before you buy.
• Drives: fit 3.5-inch SATA hard drives, up to 14 TB each. Surveillance drives are the usual choice for a recorder that writes around the clock.
• Set-up: use plug-in free access in a browser at supported versions, such as Chrome V57 or Edge V89 and above.
• POS: POS means point of sale. Till information can overlay on live view and playback, and can trigger recording and alarm.

Planning your system

The job also needs cameras, a network switch, hard drives and a monitor. Decide how many days of footage you need first, then size the drives from that.

Some functions depend on the camera, not the recorder. ANPR is by camera only, so plan a suitable camera for number plates.

Facial recognition, perimeter protection and video structuralization share the engines. Decide which matters most on your site before you commit to channel counts.

Standards and ratings explained

• ONVIF profile S/G: ONVIF is a shared standard that lets network video equipment work together. Profile S covers streaming and profile G covers recording and storage.
• Stream over TLS: TLS is the encryption used for secure web traffic. Here it protects video as it moves over the network, and playback and download can require double verification.
• ANR: Automatic Network Replenishment helps protect recordings when a camera's network link drops.
• EN 50130-4: a standard for how alarm and security equipment copes with electrical interference, part of its CE standards.