[mastermedo]

Crying wolf is the main problem. They often go off on random things. Especially outdoor sensors, as you pointed out.

But then again, one could have a short lasting warning siren, and then a full blast one when someone does break in.

[arcanemachiner]

I had to configure a sort of... inverse debounce on a cheap motion sensor I bought. It has to go off more than once (in a short time period) before the alarm is triggered.

[scottlamb]

Elk systems have a similar option called (iirc) cross monitoring. When enabled, two motion zones must trip within (e.g.) 2 minutes before the area will alarm. They could be sensors covering nearby rooms (and catch someone walking from one to the other) or have different angles of the same room. Either way, it's quite resistant to false positives.

[thomasjsn]

Agreed, I have no outdoor sensors — but will be installing an outdoor siren. It is delayed by one third of the trigger time, so with 60 seconds trigger time — the outdoor siren is delayed by 20 seconds.

Should we set off the alarm by accident, we then have 20 seconds to fix it, before all the neighbors are alerted.

[ed25519FUUU]

I found the PIR sensors to be too unreliable outdoor.

Right now I’m thinking about simply processing the outdoor cameras rtsp stream for people detection and then triggering is if it has a certain confidence.

Can be done quickly on the Pi with something like a google coral accelerator USB stick.

You can also have it send the photo it detects to your phone and you can manually turn on the alarm with that alert after viewing it.