Justin O’Bierne, a cartographer from San Francisco, has a great article about the huge distance between Google Maps, and Apple Maps. Mr. O’Beirne writes a lot about maps, especially online maps. His most recent article centers around buildings. Namely, how is it possible that Google has so many buildings on their maps, even in very small towns?
This is something that Google has been adding in the last few years. Mr. O’Beirne notes that Google isn’t just adding addressed places, but they’re adding garages and other structures as well.
Not only are they doing that, their buildings are highly detailed..
He goes on to examine a full range of buildings across the US, that show up as highly detailed models in Google Maps. He also shows that Apple and Bing have nothing even close to this imagery, so what’s going on? How are they doing it?
The answer lay in an old press release, dug up by Mr. O’Beirne. The models are coming from computer vision analysis of Google Earth satellite imagery. So, as summed up in a gif:
Not only is Google doing this, but it’s doing it FAST, and much faster than it’s competitors. As Mr. O’Beirne notes:
Just two years after it started adding them, Google already had the majority of buildings in the U.S. And now after five years, it has my rural hometown—an area it still hasn’t Street View’d (after 10+ years of Street View).
Finally, Google has also introduced another feature into Maps: Areas of Interest. Area’s of Interest are known by another name is academic research: Commercial Corridors. They’re typically defined by locations with densely packed shops and restaurants. This may seem simple, but it presents a problem to Google: how do you display all of those places on a map without the place names overlapping? If you can’t show all of the businesses, which businesses get picked? How will the user know, at a glance, which areas of the city are areas of interest?
As you can see in this screenshot of my neighborhood, Google has solved this by creating areas of lightly shaded orange.
Justin O’Beirne notes that these areas are not smoothly defined, they seem to be form by conglomerations of actual buildings, and when you zoom in, Google is actually locating where physically the businesses sit in each building.
So how do they do THAT?
Well, this post is very long now so I’ll just show you a couple more gifs that Mr. O’Beirne made:
As Mr. O’Beirne notes:
…so this makes AOIs a byproduct of byproducts:
To sum up: Google made a map of the entire Earth available on Google Maps, and then used computer vision to create detailed models of precisely located buildings. It also sent a car with a camera around the world to all the road’s that it could to give street view imagery, and then analyzed that information for signs and other details. It then combined all of that information to create precisely detailed, located buildings with precisely accurate location information for businesses and areas of interest in cities. As Mr. O’Beirne notes, Google is making data out of data.
And that’s why Google is so far ahead.