People counters are mostly used in the retail industry to gain better insights of how shoppers behave. They are also found in security, event management and smart cities applications. Imagine you manage a large mall; these counters would help you know how many people enter your mall, which paths they take, where they stop, and foremost, when does it all happen.
Just like most sensors, people counters have been around for a while. However, their data is not always centralised and connected to enterprise systems where they can support decision making. Furthermore, they can be expensive or require at least a camera and some heavy image-processing to do the task.
In this tutorial we'll explore a home-made and basic people counter (no camera needed!) and stream its data to the Ubidots cloud, where better insights can be created.
What is Ubidots?
Ubidots is a cloud service to store and analyze sensor data in real-time. It allows you to create applications for the Internet of Things, without having to be an expert in web programming, data-bases or APIs.
1. Wiring and Casing
As you can see, the motion sensor has just three pins: V+, Ground and a third one that outputs the signal: "1" when there is movement, and "0" when there isn't. No need to solder anything, nor to write I2C or serial functions to detect this signal; just plug the cables straight to the GPIO pins of your Raspberry Pi:
Here's how it looked like:
Because the sensor is very sensitive to movement, I used the jumper switch behind it to set the lowest sensibility. Also, I placed it in a dark case with a small aperture, so that the motion sensing focuses in one point instead of being so omnidirectional:
At this point, we'll assume you have done a basic setup of your Raspberry Pi and you are looking at its Linux command line. If not, we recommend going through this guide first. You can also check this post about the Raspberry Pi WiFi setup.
Let's make sure we have all the libraries:
$ sudo apt-get update $ sudo apt-get upgrade $ sudo apt-get install python-setuptools $ sudo easy_install pip $ pip install ubidots
Create a new file called "peoplecounter.py":
$ sudo nano peoplecounter.py
And write the following code into it. Make sure to replace the values of the API key and the variable ID with the ones in your personal Ubidots account.
(Note: the code is not too elegant, but hey I'm not a Python developer, just a hardware guy :)
from ubidots import ApiClient
import RPi.GPIO as GPIO
api = ApiClient("a21ebaf64e14d195c0044fcc3b9f6dab9d653af3"
people = api.get_variable("5238cec3f91b282c7357a140")
print "Couldn't connect to the API, check your Internet connection"
counter = 0
peoplecount = 0
presence = GPIO.input(7)
peoplecount += 1
presence = 0
counter += 1
counter = 0
peoplecount = 0
The script consists of a loop that checks the state of the pin #7 (the motion sensor). If it reads a "1", meaning that there was movement, then it increments the "peoplecount" variable and waits 1.5 seconds so the motion sensor goes back to normal. This is done 10 times, making sure there is at least 1 second between each cycle, then it sends the total sum of "movements" to Ubidots. If you need to calibrate the People Counter, just play with the "time.sleep" lines with other values.
When the script is finished, you're ready to run it from the console:
$ python peoplecounter.py
That's it! Here is how the raw data looked like in my case:
Now that the data is in the cloud, you can create widgets like this one to display the activity in real-time:
This project provides a hint of the amount of people passing through a particular point. It doesn't provide the exact number of people, given the limitations of the motion sensor, but in some applications this might be just enough.
More elaborate people counters use cameras and image processing algorithms to detect what the moving object is (person, car, pet..), in which direction it moves (in / out, left / right) and they could even be extended, in theory, to know the age and gender of the person.