There’s no question there’s a lot of technology going on inside Google X’s driverless cars. In fact, you can find Google patent applications dating all the way back to 2009 dealing with almost every aspect that makes their fully autonomous vehicles tick.
The latest — filed in September of last year but published only a few days ago — deals with Google’s specialty software in their driverless cars that makes it possible to scan and read traffic signals, sending that information the car or user. Google explored a variety of methods to recognize changes in traffic signal, ultimately deciding that a mounted camera would be the easiest and most logical solution. Working in conjunction with location data, these cars are able to calculate the position of traffic signals on a map using 2 or more cameras, discerning whether it’s a traffic light they’re looking at, or simply the lights of the car up ahead.
As detailed in the patent filing, the software works by first receiving the images taken by the camera, filtering images (those taken of intersections), classifying images by searching for red, yellow, and green blobs where it can then label the classified colored blobs accordingly. Using motion compensation to identify associated labels, it can determine the 3D location of traffic signals and generate traffic signals on a map.
Again, this data can be used by other driverless cars alerting users if there’s been a change (green to yellow to red). This, folks, is how your future driverless car (we’ll all be driving these, right?) will work with others to monitor traffic lights and signals, preemptively knowing when to stop or slow down. Pretty neat, right?