Detecting traffic accidents with machine learning

In several countries across the globe, governments lack advanced technological infrastructure that is required to maintain modern cities. This is especially true on maintaining safety on roads. Road accidents are an unfortunate outcome of improved roads and increasing number of vehicles. While steps should be taken to avoid them as much as possible, it is also important to have a fast-response system once an accident occurs.

Centree Technologies, an Atlanta, Georgia-based company has come up with a solution that can detect car crashes, car break-ins and other incidents and automatically report them to authorities. Centree does this using boxes of auditory sensors that can listen for these incidents, confirm them using its machine learning model and then get the police and EMS to the scene.

Elsa Perakis, Co-founder and CEO of Centree, said that all the incidents reported through their system are loaded on an interactive map platform so that users know what’s happening within a city in real-time and adjust responses accordingly.

“Our boxes include a microphone that listens for car crash and car break-in sounds, which are then run against our machine learning model that confirms them or not and finally the camera is used to provide extra certainty that an incident has occurred,” Perakis said. “In case of an incident being confirmed, it is then loaded on our map UI platform, along with information on time, location and type of incident, as well as a few seconds of camera feed and sound that confirmed that.”

Does not rely on people to implement the solution

While several solutions have focused on making roads safer, they are mostly limited to technologies that vehicle owners should make use of. What makes Centree different is that it provides safety to cities as a whole, without relying on individuals purchasing technology to install in their cars.

This particular nature of the solution helps transportation departments, police, emergency services and other agencies to have a total view of what’s happening in the city. This can reduce response time, optimize resource allocation and of course, save lives, reduce traffic and save money. In the case of car crashes, for example, getting the medical response to the scene even one minute earlier can save someone’s life, according to experts.

Smart cities to fuel demand

Perakis said that getting transport departments and cities to adopt Centree’s product appears to be the straightforward route for the company. Cities need to ensure their citizens get the fastest possible response in case of an unfortunate accident or any other type of emergency.

“The authorities want to be there at the scene right when these incidents occur, but they can't be everywhere at all times and relying on witnesses to notify them fast enough doesn't work out in a lot of cases,” Perakis said. “Our product can make EMS' work easier and more efficient, as getting to the scene even one minute earlier can help keep injuries to minimum damage and save lives. Police work will become easier as well, as we can get them to the scene faster and help them optimize the use and placement of their resources.”
 
Centree has also seen demand from tow truck companies and parking garage owners to detect car crashes and car break-ins respectively. For tow trucks, getting to the scene fast enough is important in order to get the job and clear the scene faster. In the case of parking garages, ensuring the safety of their customers and maintaining a good reputation is crucial to keep their businesses alive.

Perakis expects the developments in the field of artificial intelligence (AI) to have a major impact on their solution in the future.
“We are using machine learning to confirm the incidents right now, so as AI develops we will be able to identify them with even more accuracy and in less time,” Perakis said. “One of our next goals is to be able to predict incidents before they happen which would be a really exciting transformation for smart cities.”