Automatic passenger counting technologies (iMOVE CRC Project)

[MUSIC] The project identified and trialled a number of
technologies for automatic passenger counting. The research was a collaboration between
Swinburne’s Smart Cities Research Institute Transport for New South Wales and Sydney Trains,
and the iMOVE Cooperative Research Centre. The objective from Sydney Trains was counting the
number of passengers on the replacement bus such that they can forecast how many buses they
would need in order to provide service. Looking at how we could automate the process of
counting passengers on these buses and because you know we’re
talking about real-world buses we have to think about the real-world
conditions in which that has to work. We came up with four different options to provide and then we came together to find a way to develop
those technologies and compare them. For the vision based solution we had cameras
mounted above the doorways but we’re also looking at other vantage points
including just a bit back from the door so that we could track passengers a little bit
more easily as they entered or exited the buses. What we were looking at here was using state-of-the-art
deep neural networks as the key to our solution. Using them to detect people reliably, particularly under
the real-world conditions that we were dealing with with a bus with lots of variable lighting conditions,
vibration, lots of issues that come up which in general for computer vision are big challenges. So we were looking at how well these techniques
would work in this particular context and then seeing how we could improve on those. We’ve been working with pressure
sensors for a very long time, and we had a couple of solutions at hand which
had to be trialled beforehand in our laboratory and then we had to produce the actual prototype. We had twenty four sensors on the sensor mat which were able to tell us in the
direction in which people are going. Mobile sensing works really
well for people counting but you have to make sure that the
environment is right for it. The special challenge with us was that we
were trying this in a moving environment. We need to put our sensors all over the bus to
have some way of locating the person’s phone and then judging by how many of our sensors can
sense a certain probe request we can say it’s likely to be on the bus or
it’s unlikely to be on the bus. If you shield it appropriately
it will work much better. This sensor is particularly good
in crowded environments because by measuring the distance
you can separate different people. With a camera if people wearing the same
colours it is very difficult to separate. It guarantees privacy also because you
can’t tell whom the sensor is actually watching. The project demonstrated how we can
converge meaningful technology with infrastructure and existing assets to provide
a better user experience. Understanding how many people are using
these replacement services can provide insights especially in how the organisation can reduce costs while at the same time maintaining
high customer satisfaction. From doing the field trials we were able to
learn more about what we need to be handling for a real-world deployed system which is what we’re
looking at now is how we can further improve these techniques. We were delighted to work with partners who
share the same vision about future cities and about the role of digital innovations
in creating a sustainable world. [MUSIC]

Leave a Reply

Your email address will not be published. Required fields are marked *