I have been doing a lot of work on automation based on presence and have reached a point of minor frustration. Essentially, the options today are as follows:
IR based motion sensors
Radar based presence/motion systems
Variety of Bluetooth/WIFI/Device/GPS geographic location
With the right logic, it is possible to build a 90+% solution, but it feels like it is hard to reach the point where all members of the household are willing to rely on this for automation.
I was thinking about about an article about why Tesla was building their self-driving capabilities on top of vision instead of sensor data. This lead me to wonder is the long-term future IR of radar based or is it possible, that in the long run, enough camera's for full coverage with image processing logic that can say this many people and this many animals of the following types are in this room. I know there will always be resistance to cameras from a privacy perspective and any sort of always on will require mains derived power instead of batteries. It seems like if we reach a point where computing power is cheap enough, then a camera with recognition software might be superior to logically deriving the situation from various sensors.
Today I have a room where the lights turn on when someone enters, and then remain on as long as anyone is present. To do this requires an IR movement sensor for on (needing quick recognition) and then radar for off (needed because occupants are frequently relatively immobile).
This seems to be exactly the situation where analyzed video would be superior if it could be done locally to avoid the worst of privacy implications (such as ensuring it is not stored or sharable with unauthorized parties.
What are your thoughts. If price competitive and accurate recognition, would you replace other sensors with video based sensors?
Many people use cameras with appropriate hardware and software for motion and presence sensing. It's not cheap or plug and play and you have to understand the limitations but it works.
So effectively, if hardware advancements continue and processing power continues to get cheaper, then someday, a camera for presence sensing is likely to get into the price range of today's IR sensors. Just not any time super soon.
For all interested in DIY projects - look for ToF (Time Of Flight) sensors.
There are a couple of threads on this topic in Home Assistant forums.
I've got one commercially produced (in Korea) ToF Zigbee sensor, and so far the results are amazingly good! I have mounted it on a door frame, and I can now turn on and off the room lights almost instantly, with less than a second delay and great accuracy. No need for any PIR or mmWave radar sensors, just counting the number of people that enter or leave the room.
Sounds interesting, but until there are ToF sensors that are as available and price-competitive with the other sensors, it doesn't solve my problem.
That said, I think it will be interesting to imagine a day when both ToF and my imaginary smart vision system are available at reasonable cost. ToF has the advantage of being anonymous and not generating real images (imagine use in a bathroom), whereas a smart vision system would have the ability to not just count people, but also identify which people are there. This would be great from customizability, but bad from a privacy standpoint. That is why I would imaging the camera would have to be able to do all of the processing locally so that it would be able to put into a don't record or broadcast the image mode.
CSM-300-ZB (V2) "people counter" ToF sensor driver for Hubitat is now published here.
The only problem with this device at this time is the delivery from Korea. Hopefully, similar ToF devices will become more widely available in the next year.
