I have a room (waiting room) about 10x10 meters and I would like to locate people in there. Celling in the room is really high, so I can place the radar above in any height I want. My questions:

  1. Is it possible to use some ready/made radar/lidar solution to detect what seats in the room are currently used by people?

  2. Is it possible to detect people moving in the room on the fly (~30fps)?

  3. Is it safe to use radar on humans?

I know there are papers on this topic. However, I am not sure if there is any reasonable inexpensive solution for ordinary people. Something I can plug in PC and read raw data to process in any software I develop.

Further info:

There will be cameras. This measurement should be a complement to the camera machine vision to make it more reliable. I have experience with development of machine vision systems in difficult environments (mainly industry environment) and they have never worked as intended.

I also thought about tensometers, accelerometers, or microphones in the chairs to just locate sitting persons, but I would like to avoid collecting 20-60 channels.

  • $\begingroup$ As detail changes in the camera view the image size changes... we used that once. But are you allowed to irradiate subjects? $\endgroup$
    – Solar Mike
    Jun 18 '20 at 6:23
  • $\begingroup$ @SolarMike I edit my question. $\endgroup$
    – matousc
    Jun 18 '20 at 11:32
  • $\begingroup$ Honestly I do not deem irradiating people a good idea, especially if it's a kind of waiting room where long stay would be normal. Not to mention legal aspects, even if radiated power could be very low you should probably undergo long NHCS verifications. IMHO a second, or even more, indipendent vision system, different cameras, different view angle, different processor could be used merging and cross analysing datas $\endgroup$
    – carloc
    Jun 22 '20 at 5:09

Any reason why you want to use radar? Why not use image processing from a visible-light camera, or even an IR camera?

Edit in response to the additional details you provided: I have done some work on image comparison for intruder detection systems and is hard to get it working as intended too. Because I am running my machine vision system on platforms with low computational power, in the end I have to use very simple rule-based algorithms which often give false positives. But I need the resolution visible-light cameras give so I have to stick with them.

If you just want to locate people in the room, my suggestion would be to use IR cameras. They are not as niche as they sound - it is possible to modify commercially available cameras to infrared ones (Google it). Then you may assume that any hot spot on a seat is (probably) a person.

  • $\begingroup$ I edit my question. $\endgroup$
    – matousc
    Jun 18 '20 at 11:32

I am wondering why not just use 2 light beams at the entrances and exits to count beam interruptions. It's a simple system but effective. With 2 beams broken in sequence you can tell which direction people are moving and drowns on which beam is interrupted first whether addition or subtraction occurs in the logic counter.

The distance between the pair should not exceed a human head distance from nose to nape.

  • $\begingroup$ This too can have errors. Children being carried by parents, someone in the shadow of another, people too close, etc. $\endgroup$ Jul 1 '20 at 6:58
  • $\begingroup$ Agreed. No system is perfect. But it is preferable to high tech expensive equipment. If the beams are angled to create a crossing point in the vertical axis they may remove many errors such as your possible ones. A child in arms is difficult to detect $\endgroup$
    – Rhodie
    Jul 1 '20 at 15:14
  • $\begingroup$ Unless the camera system is very sophisticated a child in arms will be difficult to detect by vision and probably impossible with radar/lidar. $\endgroup$
    – Transistor
    Jul 26 '21 at 3:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.