How to detect the presence of a person?

Question

Smart House

A smart home is smart for that, so that he himself knows when to do what. I want to start with a simple function: turning on the light according to a simple algorithm: it is dark and there is a person in a given room. Everything seems to be clear with the light - we take the sensor and it's ready. But with the detection of a person, everything is more complicated. The popular cheap way to install a PIR detector is not suitable, because. the number of false positives is very large: they froze and it turned off, it can react to pets. You can also install an optical system of lasers that will detect the entrance and exit from the premises. But there is also a problem with false positives: turned back at the sensor level and it will consider that the entry is made. I think it would be ideal to build a thermal imaging camera and use its data to determine the silhouette of a person. But first of all, it's very expensive. and secondly, where can I get such a recognition algorithm? As a compromise, use a regular camera connected to a RaspberryPi and have OpenCV installed. But I didn’t find examples and information on recognizing people in the video. There are plenty of people, people - one article on Habré, and then with an unsuccessful result.
Are there any other options to detect a person in the room? Or does anyone have experience using OpenCV for these purposes?

Answer 1

A person usually carries a smartphone with him. The smartphone clings to wi-fi.

Answer 2

follow me and I will make you fishers of men.
There is a link to the cascade, and a phrase for Google - People Detection from OpenCV.
But I would not bother so much - a presence sensor or similar.

Answer 3

"smart home" is a system for automating processes that accompany the life of smart and not so primates. the light control function can only be implemented for a pantry or similar closets. for a living room - at the moment it is not solved by adequate methods in any form.
to begin with, it is better to make "smart" bathroom ventilation and then try to evaluate how it is better / easier / more reliable than "stupid" timed ventilation. then, in the same way, one can evaluate the potential for the implementation of ideas in a standard 50m bughouse without major repairs.