12 replies to this topic

[Michal Zimmermann]

Hi guys,
I've been looking for documents on GM technology but haven't been
lucky so far. What am I curious about?

1. How do they do it? :-) I mean the the map background (not JSON/
HTML/ etc.), how they obtain the map? It's obvious with the satellite
map, but how they make the basic map (is it called roadmap in english
version...?)? I can't imagine they vectorize satellite images using supervised classification :-)

2. How and where are all the features displayed in the map stored? In a
database? AFAIK you can't use map algebra over GM, can you? Can you
somehow interact with the basic maps, or can you call just markers,
lines and polygons inserted by users?

If anyone can help me out with these, I would be really glad.

[rudy]

They seem to rely on commercially available data sources (e.g. Navteq, TeleAtlas) for their roads and some of their other primary features. They might supplement some of their data with freely available data coming from government sources.

Not sure if they use Esri technology to produce their maps but it wouldn't surprise me if they did. in that case, the map would be built in ArcGIS, then cached or converted to a raster format. It is easier and quicker to display raster images than it is to display multiple vector layers.

[Michal Zimmermann]

rudy, thanks for the answer. I'll try to find out more about NavTeq and TeleAtlas. It's a pity there are no documents available out there about Google Maps.

[David Medeiros]

They seem to rely on commercially available data sources (e.g. Navteq, TeleAtlas) for their roads and some of their other primary features. They might supplement some of their data with freely available data coming from government sources.

Not sure if they use Esri technology to produce their maps but it wouldn't surprise me if they did. in that case, the map would be built in ArcGIS, then cached or converted to a raster format. It is easier and quicker to display raster images than it is to display multiple vector layers.

Actually I think Google switched from Navteq to TeleAtlas a while back, and then more recently began using mostly its own data. But correct me if I'm wrong.

My guess is that with the programming and engineering power Google posses they use proprietary software for most of their work. I doubt if ESRI plays any role in the production of their maps.

[Charles Syrett]

I think it's a big mishmash of data sources. I've noticed, for example, Canadian Federal data (freely available) used for drainage, as well as property lines in some areas of the USA which look suspiciously like what I see on county/municipal mapping sites. In other words, they grab whatever they can get for free, and then (presumably) replace only as needed with their own data. And clearly, they have their own brute force carto staff doing things like adding 3D buildings into downtown areas (e.g., Toronto, where they've created their own curious version of Toronto's buildings ).

Charles Syrett
Map Graphics
http://www.mapgraphics.com

[Greg Rose]

Hi,

I work for NAVTEQ and I can confirm that Google Maps does not use NAVTEQ Data. I am unsure if they still use Tele-Atlas.

From my understanding, Google uses a variety of data sources, as you will notice in the bottom of Google Earth/Maps (they cite the source at the bottom)

Google also has many employees collecting the data themselves, by driving the streets in Google cars (think Google Street View). Finally Google has implemented a way to "crowd source" their data via a tool called Google Map Maker. Crowdsourcing means that they have set up a way for the average user to update the map and create new roads, points of interest, etc. This has worked very well for Google in countries that they don't have employees on the ground. (Iran for example)

This map data is collected as vector data in a type of Geographic Information System, (Not sure of the tool they use, but I assume it is a proprietary system, vs something like ESRI ArcMap.)

Once they have collected the map data using the GIS tools, this data is put through something called a "rendering engine". This takes the raw GIS vector data (lines and points) and turns it into a raster "tile", usually in the format of .png, .jpg, etc.

When you make a call to the Google Maps, what you get back it static raster map images. (you will notice the "tiles" appear as map blocks when you move location and the new maps appear... Have you ever noticed the map appears in pieces as you pan and zoom? These are the map tiles you see.)

Once the map tiles are served up to your browser, it is not possible to interact with them, as they are just pictures.. You are correct that you can interact with the data you overlay on the map, but not the raster base map tile.

It is possible to interact with the base map and change the way the map "looks. That is where the Google Maps API comes in. By using the API, it is possible to talk to the rendering engine and tell it to do things before it spits out the map tile. For instance, you can change background colors, line widths, etc using the API, which in turn tells the rendering engine what to do. Then, the rendering engine makes the changes and spits out the final raster tile that you see at "the map".

Does this make any sense? :-)

FYI, there are actually many Open Source rendering engines out there that will take your own data (usually shapefiles) and turn them into a web based raster map. You might want to check these out, as the basic mechanism is all the same for all web based maps... GIS data in, raster tiles out.

Some of the Open Source engines you might want to check out include: OpenLayers and GeoServer.

Also, all other WMS (web map services) use this same model of rendering raster tiles from GIS data (Bing, Yahoo, Nokia, deCarta, etc.)

Hope this helps.

Disclaimer - Even though I work for NAVTEQ, all opinions are my own!

Greg Rose

[frax]

Regarding raster maps - my understanding is that the mobile Google Maps apps now use vector data as well. They have all the data there behind on the servers, even though it is not presented as vector in the web Google Maps application - that is what fuels the directions/navigation functionality.

[Michal Zimmermann]

Greg, thanks a lot for these info. The ultimate question left is: why don't they come out with some documents on google maps? Why is it such a mystery? I would so much use these in my bachelor thesis. I would also use Greg's thoughts, but as you understand that's not possible.
The API is so well documented, why can't be the technology behind the map?

Thanks a lot guys.

[frax]

Michal, I deleted one of you duplicate posts - they were awaiting moderation. I set you as validated member now, so from now on you don't have to wait for approval when posting

[Dennis McClendon]

why don't they come out with some documents on google maps? Why is it such a mystery?
The API is so well documented, why can't be the technology behind the map?

What would be the competitive advantage for Google to tell the world how their maps are compiled?

[frax]

Google has not presented much background on any of their infrastructure and services - search, maps, mail and everything. So don't hope for that. If you are serious you should try to interview someone (which I don't think would be possible...)

You are more likely to get information from ESRI on their ArcOnline maps and infrastructure, or possibly Microsoft/Bing.

[Michal Zimmermann]

Thank you all. It's obvious Google Maps work with raster data. Just to make sure I get it right: at http://gmaps-samples...zard/index.html you can style the map of your dreams, you can remove features from the map, you can change their visual representation. I guess, every option you set and confirm is sent to database and the basic db query like "if roads.visibility = off -> don't display roads at all" is done. After applying your own conditions to the map, the raster is generated from the vector database?

Am I close to the truth?

[Crackfigure]

Google maps definitely do not use ESRI technology. They are a competitor.