Virtual Earth V2 vs. V1

I’m at Microsoft Research’s Location Summit (last year I was a speaker, this year an attendee). They’ve got a variety of speakers from MSR, academia and MS product groups.

Steve Stanzel of the 3D Imagery Group (3DI) just spoke about Virtual Earth V2. They’ve had 3D data since November 2006 (Radar post). There are now nearly 300 cities are in 3D. Recently 4 cities were launched as V2 cities.


What’s the difference between a V1 and a V2 city? As I see it and understand it there are two main differences. The first is an improvement in texture quality (which makes the building look more real). In V1 the textures came from aerial imagery (the top-down view). Now they have added oblique imagery (the 45 degree, Birds-Eye view that shows the sides of buildings). The building geometry is still 2.5D, but the new textures make the buildings seem a lot more real.

The second difference is a large increase in buildings and entities in the city. The first edition of Virtual Earth featured the taller buildings in the city, ignoring most of the one and two story houses. In V2 they increased the number of buildings in a city from around 6,000 to upwards of 150,00. Most of those increases came in the suburbs. They also added some great detail work in the form of trees. In Denver they’ve rendered 300,000 trees in their exact locations. The trees come from a set of standard models. A model is placed in a specific spot based on the height and diameter of the real tree (wow).

The work that the Virtual Earth Team has done is very impressive. Much of it is built on the technology attained with the acquisition of Vexcel and Geotango. Vincent Tao, the founder of GeoTango and now a Director with the Virtual Earth team, will be speaking at Where 2.0 on May 13th.

Here’s a picture of Bali Hai in Las Vegas.

bali hai real

Here’s the V2 version of Bali Hai. Look at the trees; the rendered versions are very well-placed. As Stanzel wrote in email “we take high overlap aerial imagery and by creating a digital surface model and a classification map we are able to locate all of the trees, place each accurately and recreated actual height and diameter with a fully automated process.”

bali hai v2

tags: ,