Easy to use bidirectional people counter for one location

Question

I am looking for a bidirectional people counter for the entrance of one location to start. I am just testing some of my own ideas with the data generated by the device. The device needs the have the following traits:

• Relatively small and easy to install in an entrance
• Data is accessible live via API (wifi or 3/4G capable)
• Can count incoming and outgoing people
• Preferably overhead installation
• Will ship to the USA!

What I want is essentially this company's product: https://www.density.io/

However, I have been following them and I doubt it will ever be released. There are many other companies, but they are either offering something considerably more expensive packed with additional features or they want a monthly subscription like density per device which can get pricey. I have considered building my own device and system. Am I overlooking a good alternative?

Answer 1

Anything that involves a direction is going to require at least two sensors. I implemented something like this a while ago but we had the advantage of UHF RFID cards. We had employees with the RFID cards on them and a "choke point" - i.e. a hallway with a sensor at both ends. We could then see direction of people based on the timestamp difference of each scan.

There has been research recently into cell phones looking for WiFi connections. There are some non-trivial privacy issues associated with this approach but there are apparently companies unconcerned with the privacy issues (and I'm not sure they will ship to the U.S.) That device wouldn't really tell you a direction but a crowd count at any given time. Not sure if you could use two of them for that.

What other requirements do you have? Are these "random" people coming and going or do you have any control over them? Is this a real time requirement or can you do, for example, image recognition after the fact? Facial recognition isn't trivial yet (though it will be in the very near future) but it can be implemented even today with some pretty straight forward software.

Answer 2

An image based product is probably simpler as a stand-alone device (i.e. not integrated with an access control system). Image processing is now reaching the point where this should be something that can be implemented with a reasonably powerful SBC and a camera - so I expect a Raspberry Pi-3 plus camera might be the right point to start looking.

It sounds like you're hoping for an off-the-shelf product that you can buy in single quantities. Probably you're a bit ahead of the market here - and you're right, commercial vendors will want to sell services on top of the hardware to increase their value. This is why it's worth taking a look at what you can achieve using open software, it's likely to be much cheaper.

Answer 3

You can use infrared sensors. Put a sensor near both the in and out location—one each. Now have an Arduino board for interrupt. Whenever people enter or exit, the infrared rays will be broken and you will get an interrupt. Keep a separate count for both in and out. With this, you can calculate how many people have been inside or outside.

This is for hobby product. If you want to make a product and sell it I prefer image processing. Image processing will be easier and more reliable compared to other techniques. It's even current and future technology also.

Answer 4

I know this is an old question, but if it can help in the future:

You may want to check https://storetraffic.com, whom I work for :)

We have the PEARL People Counter sensor which is battery operated, uses Wi-Fi, bidirectional and "stick-to-install".
There is also an API available with our paid plans.