Measure people density in an outdoor area

Question

I'm currently working on a project and I need to measure how many people are in an area without any personal interaction. Basically I need to do this in parks. I don't have lot of budget for each place. All outdoor places. I want to control possible collections of people (fights or unusual collection of people).

Some alternatives include:

• WIFI Sniffing: Range 30 m.
• Physical Counter: Few meters
• Beacons: Range 6 m.

Answer 1

In a comment to the original question, the OP suggests that an accurate count of people might not be needed, but the goal is rather to be able to detect an unusual amassing of people in an area. This is a slightly different, and probably much easier, problem.

Use an infrared camera

If you don't need to count exactly how many persons are in the area, but just detect that there is an unusually large number, then infra red camera is probably your best choice.

This should limit your problem to just measuring total human-heated area.

People show up as brightish blobs in infra red.

Image source

Now - with simple image processing, we can find the pixels within a certain interval and either just add them up, to find a measure for the total area of the image that depicts something that is roughly the heat of a human. If this grows rapidly, we can suspect there is an amassing of people in the area.

Answer 2

There are plenty of methods of people counting that could be useful, depending on your exact use case.

If you want to measure the amount of people entering an area...

• you might want to consider an infrared beam counter, which simply detects when the ray is interrupted by something passing through. Here's an example which might be helpful in that case. However, it won't notice if multiple people enter at once, so it's a rough estimate.
  You've probably seen infrared motion detectors which work on a similar principle - they're frequently used on automatic doors.

• for a closed area like a building, you could attach a sensor to the door and count how many times it is opened. Of course, this will also count how many times the door is opened when people leave, but halving the number should give you a rough estimate. Obviously, for open areas, this won't work.

If you want to actually monitor how many people are in an area...

• you could use machine learning and computer vision (with a camera) to actually count how many people are in an area, as described in Lightweight people counting and localizing in indoor spaces using camera sensor nodes. This is going to be much more complex and probably more costly, but if you require accuracy, this might be the way to go. There are several commercial versions of this on the market, which you can find by searching for 'camera people counting' or something similar.

Answer 3

Well, depending on your park, you might be able to do it for free, by using Google earth or Google Maps, satellite view - although, when I looked at Central Park NY, it was empty, and Hyde Park, London wasn’t much better. YMMV for your local park, though, so it is worth looking into.

It looks like you will have to provide your own "satellite". For $50 or less, you can get a drone with Hi-def camera. The rest is left as an exercise for the reader, but I fail to see how you will get a more accurate count.

Answer 4

Is it possible to have some weight measurement plates at the entrances? With a few weight measurement plates in a line at every gate you can identify movement directions, and by using the average weight of a person, you can approximate number of people on the plate at a time. Indeed, it would be possible to differentiate between an adult and two children, perhaps.

Answer 5

Since you're not looking for a precise count, you could consider extrapolating from a smaller sample of people. Consider using a Beacon. A Beacon broadcasts a Bluetooth (BLE) signal that the people's phones pick up, and reports the signal to you over the web.

Beacons can have a fairly long range, up to 50 meters under ideal conditions. Deploying more beacons would give you even better coverage.

Not everyone is carrying a BLE-enabled phone, so you'll need to use some testing and statistics to figure out an estimate of what percentage of people do carry them. Set up the beacon, and add a video camera recording the test area. Count the number of phones responding to the beacon over a period of time, and count the number of people visible in the video. Let's say it averages out to be 20% of phones respond to beacons. So deploy the rest of your beacons, and if you get responses from 200 phones, you can estimate there are 1000 people present.

Answer 6

Kind of similar to @Aurora0001's answer, but slightly different...

In a closed room, simply have each door equipped with two or more infrared beam sensors, attached to RPi's or whatever. The first will be on the outside of the door jam, the second will be on the inside of the door jam. This will accomplish several things:

1. It will determine the direction of traffic flow.

   • If the outside infrared beam is intercepted first and then the inside infrared beam is intercepted (probably simultaneously or with a 1 second delay tolerance, depending on the distance between the beams), you know that a person went into the room.
   • If the inside infrared beam is intercepted first and then the outside infrared beam is intercepted (again, probably simultaneously or with a 1 second delay tolerance, depending on the distance between the beams), you know that a person went out of the room.
   • If just one of the beams was intercepted, you know that something intercepted one of the beams, but never crossed the other, so net change is zero.

2. If the door is narrow enough so that only one person can pass at a time comfortably, it will give you a reasonably accurate reading, because people rarely follow each other nose to toes. Obviously, the more sensors you have, the more accurate your reading can be.

I might suggest having sensors at various different heights as well, if you want to count children as well.

Alternatively, you can make yourself a weight sensitive plate which would be hooked up to a RPi or some kind of computer.