Streets Systems has been delivering equipment and traffic studies using Artificial Intelligence to analyse video since 2018. The technology and it’s acceptance as likely the best way to collect traffic data has come a long way since those early days.
For us the most interesting aspect of the technology has been it’s flexibility and the ability to drill down into real and perceived problems on the highway. One of our first projects looked at the influence of pedestrian density and effectiveness of targeted interventions on cycling speed for City of London. There’s still, as far as we know, been nothing as sophisticated done to look at friction in shared spaces since.
Over time we’ve learned that one of the most powerful applications of this technology is to find and extract instances of when something unusual or risky (or possibly risky) happens on the highway so examples can be reviewed and understood.
One example of this is pedestrian – vehicle interactions at zebra crossings. There may be a public perception that near-miss incidents are occurring, but are they? After video footage has been processed using AI it’s then possible to extract all instances of a specified interaction happening. So in the case of a zebra crossings that might be all instances where a vehicle crosses a give way line within a few seconds of a pedestrian or cyclist moving over a kerb line.
We know from work we’ve done on some crossings that there can be a higher rate of near-miss incidents for both runners and cyclists than for slower pedestrians. This suggests that approach speed may be a factor but there is a lot more that could be done to drill down in to this. A logical next step would be to test interventions which might reduce the prevalence of this friction between users.
The same methodology can be applied to any other type of interaction, so you might want to look at cases where buses pass in close proximity to pedestrians, or trucks near bicycles etc.
From a road safety perspective this approach gives you the ability to look at a location where collisions have occurred, then review all cases where a similar collision “nearly” occurred. The severity of these near miss incidents is best assessed by a human being, but finding them is perfectly suited to AI software.
From a privacy perspective there is a limit to how much a public space should be filmed in order to investigate a problem. The privacy impact needs to be proportionate to the problem. A solution to that is to monitor locations for longer periods but to only capture footage when some of the risk criteria is met.
There’s no doubt at Streets Systems that the role of AI in monitoring the highway will continue to grow, and it won’t just be about counting cars. Much more sophisticated use cases are emerging that can contribute towards a reduction in risk in particular for walking and cycling.
Get in touch if you’re interested in hearing more about how this can work for you.
