Road accidents and risk prevention: Using Lanternn by Valerann™ for making roads safer

OVERVIEW

Most new vehicles today – whether commercial or private – are fitted with advanced vision sensors that aid with areas such as hazard detection and driver assists. In this case study, we look at how Valerann® has been able to use vehicle vision systems as a data source for Lanternn by Valerann™ and with AI-powered analytics has been able to identify temporary but highly risky events taking place along highways, including a real-world example on one of Israel’s biggest highways.

THE CHALLENGE

Different data sources have different values. Stationary cameras, the most commonly used road monitoring technology, provide a view of the road but they can only detect events in their line of sight, which is also the same for other sensors such as radar. Crowdsourcing, a newer source of data to monitor road conditions, is also very useful for getting real-time on-the-ground information, but reporting is not uniformly done and information is not always accurate.

Valerann®’s speciality is understanding the unique value of different types of data in order to cut out the noise and provide actionable insights to road operators. Lanternn by Valerann™ is a powerful tool that fuses data from a number of traffic sources and its open ecosystem means that more data sources can be added over time to provide an even richer picture for operators.

A new data source that has been able to provide invaluable insights to operators is vision sensors – also known as dash cams – fitted to vehicles, which are now standard on most new cars to aid with hazard detection and driver assist functionality. Working with one of the biggest driver-assist technology providers, Valerann® has been able to prove this concept using vehicle-generated data from the Ayalon Highway in Tel Aviv, Israel, one of the busiest roads in the country.

The Ayalon Highway hosts a quarter of a million journeys daily. Due to this high use intensity, it's normal that many types of events will occur on the road – in fact, Valerann®’s work monitoring and analysing the highway has shown there can be as many 1,000 in a single day.

With so many incidents happening, it is impossible for the operator to deal with all of them. They instead must take a proactive and preventative approach to managing the highway and prioritize only the most important events.

THE SOLUTION

Most new cars are now fitted with driver assist technologies as standard, improving safety for drivers, passengers and pedestrians through monitoring of the road and taking actions quicker than a human is capable of. These driver assist technologies include dash cams that record video footage when the car is turned on and in motion.

For this project, Valerann® was able to obtain real-time GDPR compliant access to dashcam footage of drivers using the Ayalon Highway.

The AI-equipped dash cams would automatically send Valerann® information on anomalies occurring on the road. These often fell under the ‘Road Safety Alert’ class such as a stationary car, a pothole, or a pedestrian spotted on the road.

After analysing the data using its advanced AI and ML algorithms, Valerann® noticed the dashcams had a higher accuracy and detail when it came to identifying risks and certain patterns that were not captured by stationary cameras and open sources. In particular, incidents or events that can be classified as ‘temporary, but high risk’ such as near misses or a pedestrian crossing the road.

More precisely, vehicle-generated vision data revealed the repeated occurrence of pedestrians moving across the highway at certain locations, putting themselves at risk of a collision.

To find out more, using Valerann®’s data fusion technologies, a map was created of all the hotspots where pedestrians were spotted doing this.

This map was then cross correlated with Valerann®’s machine vision and video processing capabilities – as well as other data sources, including IoT sensors, connected vehicle data, stationary cameras, navigation apps – to determine what was actually happening.

The analysis revealed that people were using motorway exits as a shortcut to their destination. Cars were dropping pedestrians off at certain exit slip roads, where they would then jump over the safety barrier and head off to where they wanted to go. In one location this was, somewhat ironically, a cemetery.

These temporary but highly risky behaviors, which could easily result in a major accident, were not picked up by normal sources of road monitoring because they were fleeting. It was only possible to discover them thanks to the real-time dashcam footage – essentially a car passing through the entire road filming – and the subsequent analysis by Valerann®’s data fusion engine.

It also showed that dashcam footage could be useful for validating accidents reported on Waze or determining their cause using historical analysis.

CONCLUSION

If an operator doesn’t know that this risky behavior is happening, then they cannot take action to stop it.

Once equipped with this detailed information they can take preventative measures, such as: putting up warning signs to deter risky actions; placing a camera at the hotspots to monitor events further; strategically place patrols, or block the shortcut with a fence placed behind the safety barrier. These actions will then go some way to removing this risk altogether and significantly increasing safety.

As this case study shows, Valerann®, through its data fusion techniques, was able to understand the value that vehicle vision systems and dashcam footage could provide to the road operator, which was uncovering these highly temporary but very risky events that other data sources did not pick up because of their limitations.

For a highway as busy as Ayalon, using different data sources to expose blind spots in monitoring – while not inundating the operator with information that isn’t useful – can help them proactively manage the highway and help put measures in place that makes highways safer for all.