12

Inside Google Street View, moving the mouse around over different buildings and stuff, it highlights the 3D plane in which that surface of the building is located.

How does it recognize this thing? Is it done automatically by machine algorithms, or manually in the preprocessing?

Iulius Curt
  • 4,984
  • 4
  • 31
  • 55
  • I've often wondered the same thing myself, and would love to know the answer to this. – nageeb Jun 16 '11 at 06:47
  • 1
    I'll avoid posting this as an answer, as I have no clue exactly what they do, but I'd guess it's something similar to this: http://www.fieldrobotics.org/~cgeyer/OMNIVIS05/final/Bonarini.pdf Don't the street view cars use a similar camera + conical mirror setup? – Joe Kington Jun 17 '11 at 06:07

2 Answers2

4

Read this to get your answer. What Google does is it gets a panoramic view by stitching multiple images together, and then the device measures the relative distance of every object around it (through lasers), and constructs a 3D model of the surroundings based on that.

Kunal Vyas
  • 1,499
  • 1
  • 23
  • 40
2

You want to look into space-filling-curves. A sfc reduce the 3d complexity to a 1d complexity. A sfc subdivide the 3d space into 8 tiles thus it resemble an octree.

Micromega
  • 12,486
  • 7
  • 35
  • 72