Last night, the San Jose Mercury News reported (registration required) that Google is licensing technology from Stanford University which will enable it to produce 3D building models automatically. The technology comes from robotic car technologies developed by Stanford which won the 2005 DARPA Grand Challenge. In that race, the Stanford robotic car drove 131 miles through a New Mexico desert autonomously. The technology will enable Google to capture shapes and photos of buildings and allow them to automatically generate 3D building models which can be viewed with Google Earth. A Google spokesperson said they will make an announcement at Where 2.0 week after next. I’ll make sure to report on this, since I’ll be attending Where.
Google’s approach so far has been to use its popular, and very powerful, 3D modeling software SketchUp to generate quality models. The resulting models can look stunning inside Google Earth when built properly. Google created a 3DWarehouse where anyone can upload models. And, there are now thousands of buildings from cities all around the world viewable in Google Earth as a result. However, there are some drawbacks to this approach: 1) the models are built by many people and the quality and level of detail varies a great deal; 2) only the buildings which are more popular or of personal interest get attention – all the rest of the buildings of a city generally are not built.
Meanwhile Microsoft acquired technology for automatically generating 3D models of buildings based on photos taken from the air from multiple angles. Already, Microsoft has released about 50 cities where a substantial portion of all of the buildings are modeled with phototexture details. These are visible with their Virtual Earth 3D plug-in available for Windows XP SP2 or Vista with IE or Firefox. The VE 3D building models look quite good from a distance, but the quality of the 3D models released so far has been lacking upon closer inspection. Some of the VE 3D buildings are hand-made, so the quality is better in some of those cases.
Google will probably take the approach of continuing to use quality hand-made models and the new automatically generated models will fill in the gaps for other buildings. I’m looking forward to whether they will show examples of the new technology’s capabilities.
Here’s a picture of the Stanford car shown by Google during the 2006 CES Keynote by Google co-founder Larry Page.
(image from Engadget – Jan 2006 CES Google Keynote)
Read my earlier thoughts on the comparison between Virtual Earth and Google Earth 3D Buildings.
Story via TechCrunch and a tip by Mickey of Gearthhacks.