Public transport could be the first to benefit from AI
In a previous article we saw how Europe is at the forefront of the development of AI in transport, and in particular in the creation of autonomous transport. This October, another innovative new project was announced in France by the RATP, which decided to launch an autonomous bus line in 2019. The tests will be carried out in the Val de Marne on line 393, with passengers on board, and may go as far as full automation. But while autonomous transport makes a lot of noise, it is far from being the only revolutionary high value-added project that AI could bring to the public transport sector. This sector alone contains a large number of possible applications for AI and visual recognition in particular, making it potentially the primary beneficiary of AI.
To give you an idea of the wide range of possible applications of visual recognition in the public transport sector, here are some examples of concrete use cases.
Station security enhancement
Let’s start with security and safety applications. By equipping existing video surveillance cameras with an automatic detection system, it is possible to prevent many accidents and restore confidence to users whose sense of safety has deteriorated. Indeed, a study published in January 2018 by the Observatoire national de la délinquance et des réponses pénales (ONDRP) shows that more than half of women (51%) and 38% of men report feeling insecure when using public transport. An AI trained to recognize a potential danger (an assault, abandoned baggage) can alert a security guard in real time who can immediately respond on site. Similarly, real-time detection of people falling on the tracks or fainting would make it possible to rescue them as quickly as possible and automatically stop a vehicle in case of danger.
Passenger flow management
It should also be noted that cameras equipped with visual recognition can accurately count users at stations, on platforms and in vehicles, thus optimizing the management of passenger flows and the hourly frequency of trains. They provide a better understanding of what is happening inside public transport, collecting data on empty seats, off-peak hours, but also discomfort and incivility on some networks. These valuable indicators allow companies to quickly adapt to conditions on the ground.
In stations, visual recognition makes it possible to detect and estimate fraud – a major loss of revenue for transport companies. Indeed, according to the Court of Auditors’ estimates in 2016, fraud-related losses represent a loss of revenue of €191 million for the RATP and the SNCF. Thanks to smart cameras, it would be possible to automatically detect people, for example, who straddle the access turnstile and do not stamp their tickets.
In addition, the entire maintenance sector can also benefit from video recognition. Indeed, smart cameras on the tracks can detect problems such as missing screws, cracks on the rails, worn pantographs, dirt or graffiti on trains. This information goes directly to the technicians in charge of maintenance, who can then act effectively as quickly as possible.
Improved customer service on board
Finally, there are other possible applications of AI in the public transport sector, in areas that may be less expected but are just as interesting for companies, such as in-train catering for example. TGVs could be equipped with automatic cash registers that detect food and drinks on passengers’ trays, making the queues at the bar more fluid and helping to make the traveller’s experience more pleasant.
We can therefore see that the public transport industry aggregates several sectors (safety, mobility, industry and catering), making it a real attractor of use cases for visual recognition. The good news is that their applications can be put into production as early as 2019 because they are mainly based on the operation of existing video surveillance camera networks. This not only allows for accelerated implementation and scalability, but also better amortization of the existing system. Production start-up costs are low due to the reuse of an existing camera fleet, and algorithm development time is reduced due to data availability. Indeed, since the data emitted by video surveillance cameras is already collected to train detection algorithms, it is possible to develop visual recognition systems in only a few weeks, and then to increase performance over time in the field. The profits for public transport, if they acquire AI, can therefore be made in the short term. In addition, the return on investment is very high because the value provided is identifiable and the benefits are multiple.
We are already seeing that public transport industry players are taking advantage of this golden opportunity, as is the case with the RATP’s autonomous buses, or the SNCF, which recently launched a call for tenders for a video protection project. France is on track for industrial projects led by AI to be launched in 2019, for the greater benefit of all citizens!
Thanks for reading!
Augustin Marty, CEO @Deepomatic