Sometimes even the most capable electronics teams can benefit from a little external help.
In order to highlight and target passenger safety and security on public transport, a leading transport systems company asked us to find a better way to track passenger movements in and out of train and underground carriages. Existing methods, based on the use of ceiling-mounted cameras, suffered from poor detection rates in low light conditions. Poor camera angles on underground trains due to the limited head height available were also a problem.
Our customer wanted a solution that was highly accurate, compact and unobtrusive, and able to count multiple passengers simultaneously moving in and out of carriage doors. The system needed to overcome potentially confusing influences such as large items of luggage and passengers standing in doorways.
Our initial work explored the reasons why existing systems were not sufficiently reliable and examined potential installation options. We concluded that a combination of ultrasonic and infrared proximity sensors, mounted at the top of the train door opening, would be the best choice. A suitable array of such sensors gave us the ability to discriminate between outward and inward movement, between humans and luggage, and to track multiple passengers simultaneously.
We then devised and experimented with detection algorithms to process the sensor data and generate passenger counts, testing at each stage using a full-size rig in our lab. Using multiple volunteers, we recreated a wide range of scenarios, evaluating and improving our algorithms at each stage.
We first developed and optimised the necessary algorithms and produced a demonstration of how the customised technology would work. Downward-pointing rows of sensor modules above or next to the doors were installed to count the number of passengers as they entered.
A range of different ultrasonic and infrared modules were tested. The performance was then evaluated in real-world transport environments to see if there was any discrimination between the recording of people and objects.
The materials and manufacturing costs of the sensor assemblies were then value engineered and investigated to see if the project was viable.
We rapidly delivered a prototype design to our customer for which we had demonstrated better performance than existing camera-based solutions could achieve. We also supplied our test rig so they could perform their own evaluations and transferred of all the intellectual property, so they had ownership of the design.
Our proposed design exceeded the customer’s counting accuracy target, was suitable for installation in a wide range of train and underground carriage doorways and had a low component cost.
Armed with a concept for an entirely new approach to the problem, our customer took over the development at this point, using their own highly capable electronics design and manufacturing teams. In a few short months we had been able to bring some original thinking to the problem, with a fresh perspective that helped our customer explore a wider range of routes forward.