Maps have existed for thousands of years and are still evolving to continue meeting users’ diverse and changing needs. Part of this evolution is the level of details included in the maps; geometry and street names alone are no longer sufficient, and in recent years, additional attribution has been added to maps to enhance the user experience. To facilitate this attribution, TomTom is developing new, innovative tools, such as Artificial Intelligence to speed up the process of map development and updates.

At TomTom, we have already started this with Traffic signs. They provide critical information to drivers and vehicles, from warning them that they are going above the speed limit or approaching a dangerous curve to providing an environmental context for autonomous cars to plan their next manoeuvre. Capturing and maintaining these traffic signs to include them in our digital map database used to be done almost entirely manually, with operators browsing through thousands of hours of mobile mapping imagery to find and identify relevant signs.

With the introduction of laser radars in our mobile mapping fleet a few years ago, TomTom took the first step towards automating the detection of traffic signs. Indeed, sign reflectivity now allowed us to easily filter out images containing relevant sign content. Once the relevant sign content was singled out, sign classification, the complex process of identifying the category and type of sign, would occur. The human brain in naturally attuned for this complex visual process – but machines had to be taught.

Over time, our operators have classified almost 100 million signs across more than 100 traffic sign categories, creating a huge traffic sign evidence database. Taking it all a step further and through the supervised learning of deep neural networks on this traffic sign evidence database, TomTom developed its own Traffic Sign Classifier Artificial Intelligence. This AI tool allows TomTom to classify detected signs in an automated way with very high accuracy levels. Where needed, unclassified signs are revised by human operators, further improving the learning and competence of the Traffic Sign Classifier AI.

Once traffic signs are classified by the AI, TomTom’s map fusion processes kick in, adding the correct signage data to the appropriate TomTom map layers, including our navigation and HD maps.

With the explosion of sensor equipped vehicles, traffic signs will no longer only be detected by TomTom’s mobile mapping fleet but also, increasingly, by regular cars continuously driving the roads.

AI is set to play a crucial role in the quick and efficient processing of data, therefore enabling TomTom to bring the most up-to-date maps to its users for a safer and more comfortable driving experience and also, one day, for autonomous cars to make their own decisions.