One and a half metres: are you standing too close?
How well do people really follow the 1.5 metre rule? That was a question preoccupying Niels van den Nieuwendijk and Timo Duindam. At the peak of the corona pandemic, the two Engineering Physics students followed a sensor technology minor. They developed a system that registers how well people keep the desired distance at busy places, such as a shopping street or bar.
Combining sensors with physics
During their minor semester, Timo and Niels wanted to delve deeper into sensors and big data to make things smarter. They wanted to learn what was behind the sensor component and add physics to that. As the two have known each other since primary school and like working together during their studies, they decided to enrol together for the minor at the University of Applied Sciences Leiden. In the second part of the minor, they worked with Dirk Jan Remmerswaal, a Computer Science student, on their own project.
Determining depth with two cameras
Corona dominated the media, so the group thought up a device that measures distance and detects when people bump into one another. Working with cameras turned out to be a cheap and sufficiently accurate way to determine depth. I use plural for good reason because one camera alone cannot detect depth. Niels: ‘We solved that by using two cameras. You use the difference between the two images to see the depth. That’s called stereo vision.’ They wanted to find out more about it.
Just put a plug in the socket
Niels: ‘It is a matter of precision to set up the position of the cameras and it takes a lot of computer power. You must do a lot of calibrating. If a camera is tilted by even half a degree, the measurements will be wrong.’ To avoid having to calibrate them each time in every bar, they decided to attach the two cameras close together in a plastic casing. All the bar boss has to do is screw the white box to the wall and put the plug in the socket. Timo: ‘The entire minor was focused on being able to easily expand the system. The architecture behind it had to be scalable.’
Seven euros material costs at most
There are various ways to achieve image recognition, but the product had to be developed in a short time, be easy to instal and affordable. Whereas other sensors require something under the ground, with a camera all you need is a socket. The plastic box containing the DIMOD (distance monitoring device) comes from a 3D printer and cost 30 eurocents at most. You can buy the two camera modules directly from the factory in China for five euros. All in all, no more than seven euros in material costs.
‘The statistics had not been done before.
I like to be busy
with things you can’t google’
During the project, Niels, Timo and Jan Dirk focused on the statistics component. That hadn’t been done before and they enjoyed it. It included a lot of advanced programming work. The human recognition that the camera registers is calculated by a local computer and subsequently sent to a national server for storage. Customers can then view the camera images and statistics from the previous days on the website using their phone.
Coloured circles surrounding people
If there had been more time, the students would have liked to have done something with awareness. Niels: ‘We had some really great ideas for that. For example, using a laser projector or beamer to draw 1.5 metre circles around people and changing the colour if people got too close to each other.’ And then make that visible on a big screen. They didn’t have time to develop that. However, there is a counter on the bar with pointers that indicate how many visitors there are and how many times people bump into one another. But to be fair, this was mainly a creative way to incorporate the mandatory components from a robot kit in their product.
Robot takes a plant to the perfect spot
The minor supervisor was impressed with the level of work by Timo, Niels and Dirk Jan. Patrick Morley: ‘They successfully applied the theory to a current problem, did a thorough analysis and delivered a working prototype.’ None of the other students worked with cameras. What did they do? Timo: ‘Most were busy with plants. One group put a plant on a robot, and it drove through the room to search for the ideal conditions for the plant. It would also measure the humidity and give off a signal when the plant needed watering.’
Learning to use Python in a new way
Our students are also extremely pleased with what their minor produced. Niels: ‘We had learned the basis of the Python programming language during physics. Only there it was about data analysis. Then you are talking about a one-off calculation. This was about real-time human recognition. We had to write codes that run continuously and collect new data. It means you encounter other issues. If there is an error message, you don’t want the entire system to crash but you want a message stating which part of the system isn’t working. I learned a lot, particularly when it came to programming.’
Making a system scalable
Tim adds: ‘Above all, we had a great time. We worked enthusiastically on the project. The fact the scalability was so significant, forced us to think differently. It had to become a layered system. We use problem-solving skills during the degree programme: looking how to solve something when things aren’t working. That was useful in this situation too. It doesn’t always go as planned. But when you see progression in something, it encourages you to continue. I even got an internship out of this minor. I am now doing an internship at the company of one of the guest lecturers.’