eTech 2007: Session: Microsoft Virtual Earth
John Curlander, general manager of Microsoft Boulder talks about The Making of Virtual Earth
Microsoft Vision (envisioned by Bill Gates)
- Create realistic, global 3D framework
- Foundation for the 3D web
- Core of entertainment, commerce, information
In the future all website will move to a 3D experience he says.
Virtual Earth 3D was born. Microsoft approaches Vexcel in late summer 2005 and wants to do 3000 cities in 5 years (and I saw Portugal in one of the slides). Wide area coverage with automated processing was the only solution. Shows a making-of video of some sort.
Microsoft and Vexcel built a special 220 Megapixel camera for the process, the Ultracam.
Shows the “Syntopic” technique they’re using to do aerial photos of the land: Position of the four camera cones at identical positions (Syntopic exposure)
Requirements for the solution were:
High information
- Capable of >400MB/sec
- Equivalent of 10 ikonos of Quickbird satellites
- Focus on VE AOIs
Key to automation in sensor and collection methodology
- Multi-spectral imaging
- Hight overlap images, many looks at target
- Hight dynamic range, 12+ bits
Typically on a city they get 90% forward overlap with this camera. Shows the color film and the dynamic range they get with a lot of detail: 5-6 bits.
Microsoft only collects and processes data, they hire commissioned flyers to do the shooting.
Virtual Earth adds live services to the imagery data like live traffic, etc.
Workflow:
- Raw image ingest
- Initial classification
- Image orientation
- Dense matching
- Orthorectification
- Refined classification
- Bare earth topography
- Building extraction and texturing
Shows up detailed explanations on each of these bullet points and a live-demo on some of the cities.
Challenges with automation:
- World is 3D (Multiple z values for earch x,y)
- Trees: Tree detection
- Too manu facets
Look to the future
- Shift emphasis to streetside processing: true 3D
- Tree rendering
- Improved community tools
- Interiors
Labs: Photosynt, Tree detection and Image-on close range photogrammetry were mentioned.
Wraps it up with a video of New York with the obligatory Frank Sinatra soundtrack. Lots of questions after, many concerning privacy issues due to the increasing detail in photos we’re having.
Recent Comments