City of Szeged Uses Cameras, Wi-Fi and Weight Measurement To See Where and How City Dwellers Travel

In one year, the SASMob project will reach the end of its rollout period. One aim of the programme is for the city of Szeged and Szeged Transport Company (“SZKT”) to see what means of public transport people use and how. Traffic count covers intracity transport, i.e. data are collected about traffic volumes in both directions on the City Centre Bridge. IT experts can tell if a passenger getting off a mass transportation vehicle continues his/her trip from the same stop on another vehicle. All this is examined and counted fully anonymously. The goals are to improve the quality and efficiency of transport and to reduce traffic jams.

If you want to get a picture of a city’s transport, you must examine both public and private transport. In the SASMob project, Szeged puts a major emphasis on monitoring public transport. This includes a number of aspects, the first being a “static picture”, i.e. watching a certain point and counting how many people travel through it. Such counting points, or gates, have been installed on the City Centre Bridge, to count the number of passenger cars, bicycles and trucks travelling across the bridge. A similarly important goal is to count the number of passengers getting on and off a given public transport vehicle.

– These items of data are important for Szeged and Szeged Transport Company, but we have gone further: these trips have been organised into flows – said Vilmos Bilicki, Assistant Professor at the Software Development Department of the Institute of Informatics, Faculty of Informatics of Szeged University. – As an example, we want to see which transport nodes along the route of a trolleybus were the ones between which there was a high number of passengers. We started to organise community transport data into flows on the route of trolleybus 8. It was then that it turned out, for instance, that on a given day on a given trolleybus many people travelled from the stop in Gém Street to Dugonics Square. With this point-like measurement we could demonstrate how we can organise trips into a system. The first-level connection that can be seen though our measurements is the two points – the starting point and the destination – between which there were the most passengers on a given service.

The IT experts working in the project can perform these counts so that the results are fully anonymous. – We are interested only in the passenger numbers, how many got on a trolleybus in a stop, and how many got off it in another – added the Assistant Professor. Vilmos Bilicki then spoke about a higher level of aggregation, when the survey focuses not only on a given vehicle and the passenger traffic in it. – So we know that, say, twenty inhabitants of the city got off the trolleybus in Dugonics Square – but, if we are to organise trips into flows, we must also know what happened in that bus stop after leaving the vehicle: for example, people started to walk, or four of them continued their travel on another service. The data collected so far are extremely useful for the organisers of public transport, but these additional items of data are even more important.

The next, even higher, level of traffic count is measuring the entire city’s traffic and organising it into such flows. – For the first two measurements, there are sensors installed in the community transport vehicles, which can provide us with this data. Counting at points is used for city-level traffic organisation: the gates installed on the City Centre Bridge and mentioned above are suitable for these measurements. If you create a system of such gates and connect it to community transport (and now I don’t want to go into the details of the principles of counting), you can create such flows at the city level. The program developed by us offers this feature and can also follow the transfers mentioned before – said Vilmos Bilicki.

There are a number of good solutions and methods to count traffic and, as part of that, participants in community transport. The development work of the Software Development Department focuses on the protection of personality rights, which enables them to create flows – which is by no means trivial. – We can create flows by not having numbers at certain points but between different points, i.e. there is a tracking method. You can track an individual, a vehicle or even a mobile phone carried by someone. The latter is possible with the help of Wi-Fi. The chart prepared about passenger traffic on trolley bus 8, which I mentioned earlier, was made with this method. The phone has an ID, which is anonymised by our software, i.e. encoded with an encoding method that guarantees that the input cannot be traced back from the end result, yet the data is clear. We use extract functions and group IDs. Obviously, this method also has its limitations: today, 60-70 percent of mobile phones always “ring” at a different address. Therefore, this passive counting method currently relies on only 30-40 percent of mobile phones, and we expect no help from either the phone or its user with tracking. As a result, we can only tell orders of magnitude, the system has a level of uncertainty. This is what we have developed so far, it works in a stabile way and can be used for several more years, offering valid information. Besides, it is very cheap. However, we are continuously working on its further development – said Vilmos Bilicki.

Today, three different measurement methods are used on community transport vehicles, one is weight measurement. There are vehicles which measure their own weight, and, with this method, we can tell with an accuracy of ninety-three percent how many passengers are travelling on a given vehicle. This information can be used to see passenger flow on, for instance, a monthly basis. For this, the Institute of Informatics has pretty good algorithms. – We consider that as the basis, on which we build camera and Wi-Fi based counting. With the latter, we have an accuracy of over eighty percent, which means it is fairly reliable, and this is what we are developing to track transfers – said the Assistant Professor. Cameras can be installed even at each door, to be able to count get-ons and get-offs. Again, the individual is of no interest to us – the only thing that matters is how many people got on and off the vehicle at a given stop. This heterogeneous counting system, installed on individual vehicles, improves its own accuracy. Flows can be generated using multiple counting methods – which later enables the counting of passengers by, for instance, age, i.e. how many students were travelling on a given service.

“Normal” traffic count was done with the traffic counter gate installed on the City Centre Bridge – here, the next step is to start monitoring passengers who join public transport at the two ends of the bridge. – The city of Szeged now wants us to install these counting devices on the vehicle pool of Szeged Transport Company. For the time being, the buses operated by Volánbusz have no such devices but, to my knowledge, the city management are having negotiations with this public transport company also about installing the devices, which are used by SZKT and work well, also on buses. As a first step, a representative sample of these buses must be created, on which these counting devices can be used. They will be installed on the vehicle pool of SZKT in the first half of the year, simultaneously with the installation of the traffic counting system on the City Centre Bridge. Once installation is finished on the vehicles of SZKT and continuous counting begins, a number of questions can be answered based on these data, like which the busiest periods of the day are or what traffic was like at the weekend – said Vilmos Bilicki.

There is still a lot to be developed in the system. IT experts always perform these development tasks taking into account the protection of personality rights. It would be simple to just use a camera on the bridge to take photos of number plates, but this is not the aim. On the contrary. Our research activities are aimed at not even trying to read the plate number from the recorded picture but, let’s say, we create a string of numbers about the back of the car, based on the features of the picture. This string is generated by an algorithm that uses artificial intelligence (adapting a face recognition algorithm to the backs of cars), and although the plate number is also likely to be used in the process, it cannot be traced back from the number string. Here, there is also a possibility of mistakes, as the system can confuse two red cars of the same type. These pictures can also be used to create city-level traffic flows – these are the things we are working on now. We have one year left of the project rollout period, by the end of which current developments must work flawlessly. However, the system can be further developed afterwards. Just to mention a fairly simple example: one goal can be to be able to differentiate trucks by size. Speed measurement is a similarly important aspect, which can be applied to cars and bicycles alike – and, eventually, giving a forecast of the length of traffic jams. That the system will operate securely and that the infrastructure and software will be available can be taken from granted. The program is not Szeged-specific: it can be used anywhere. Another priority aspect was low price – added Vilmos Bilicki.

The Assistant Professor added that, besides these, they have also started to measure air pollution. They have prepared a sensor, which is being certified. – These can be installed on the top of trolleybuses and can measure 1-, 2.5- and 10-micron particles. This information will draw us a relatively precise map of air quality about the parts of the city covered by the trolleybus service – concluded the Assistant Professor.

Garai Szakács, László