Cameras are the black sheep of IoT

Cameras are a little bit different to other IoT devices. Here’s why.

Eyeballs v Algorithms

With CCTV cameras, the output is images. Don’t bother trying to get data from them, it’s expensive and disappointing.

Instead, use images for what they’re good at — transmitting information into the human brain.

Let the IoT device trigger the event, then look at the images.

That’s the root of the issue, now here are some details:

Connection Protocols

Unlike every other IoT device, high bandwidth requirements mean that cameras don’t use Bluetooth or Zigbee. We even advise customers to avoid WiFi.

Communication Protocols

This is not a place for MQTT or ZeroMQ. Like with the connection protocols cameras are often sending in the region of 1mb/s streaming video and so the protocols designed for small amounts of intermittent data aren’t appropriate.

Cloud v Edge

It’s easy & cheap enough to take the output of a temperature sensor, polling every minute, and store it for almost infinity. With live streaming video, that would start to get expensive. As the survivor of a cloud CCTV business (Camba.tv) I learnt the hard way that it makes more sense to store your video locally, if you can. Now, with 64Gb SD cards on the camera, you can.

Privacy & Data Protection

I can store year’s worth of air quality readings or voltage measurements without worrying too mcuh about personal privacy infringements. A jpeg containing a face or a number plate is quite a different story.

Impact

Sometimes, this: Tide Status: 1.3 just doesn’t cut the mustard quite like this: