More analysis of 3D Imagery

When Google first announced the automatically generated 3D imagery, they said it was generated via stereophotogrammetry from aerial imagery. In our recent post about 3D imagery featuring cruise ships, we explained that it involves taking pictures from different angles and using that to automatically generate the 3D structure. The result of this is that moving objects cannot be captured by this technique. To understand the implications, lets have a look at a few examples around Charles de Gaulle Airport near Paris, France.

Aircraft moving and stationary
Above left: A moving aircraft on the runway has no 3D at all. Above right: A stationary aircraft is in 3D.

Ghostly aircraft
When the aircraft moved during image capturing, we get ghostly effects.

The technique used for creating the 3D does not handle over hangs at all well, which is one reason for even stationary aircraft not looking very good. It is also very noticeable on bridges, and the way trees tend to look like bushes with vertical sides rather than a trunk with overhanging branches.

Tower
This water tower demonstrates the problem with overhangs. Above left: Water tower in Street View. Above right: Water tower in 3D.

The water tower above shows six distinct images were used to create the faces of the tower, and possibly a seventh for the top. Two of the faces have much greener grass, and the towers shadow is in a completely different direction so they were taken at a different time of day, from the other four faces.

We have put all the above locations into a KML file. It also includes a number of other notable locations around the world, demonstrating that for intricate structures, Google often manually improves on the models. This is one reason why certain locations take longer to release than others. Bridges, particularly, seem to get a lot of attention, so cities with a large river and lots of bridges can expect to take longer than others.

Tilt_RotateButtons
For more interesting effects, look at one of the above aircraft in Google Maps Earth View by clicking here then try tilting the view and watch the aircraft disappear. Next, rotate the view and see how the aircraft is only partly there.

Also have a look at this aircraft. Note how the same aircraft is parked there in overhead view and the tilted view, but as you rotate, the baggage trucks are in different locations.

If you do not see the compass and ’tilt’ buttons in Google Maps, then it is most likely that you are in Maps’ Lite mode. You can find the minimum system requirements for seeing 3D in Google Maps here.

About Timothy Whitehead

Timothy has been using Google Earth since 2004 when it was still called Keyhole before it was renamed Google Earth in 2005 and has been a huge fan ever since. He is a programmer working for Red Wing Aerobatx and lives in Cape Town, South Africa.






PLEASE NOTE: Google Earth Blog is no longer writing regular posts. As a result, we are not accepting new comments or questions about Google Earth. If you have a question, use the official Google Earth and Maps Forums or the Google Earth Community Forums.

Comments

  1. problem is that they use footage coming from planes. thats why trees usually have no references for the lower bodys, since that area cant be seen in the camera. still the generated mesh is the best available datasource and i would love if it would be available for the public (it would help a lot with architectural projects and so on)

  2. Bridges…so that is why we are waiting so long for big cities like Budapest?

  3. I’m glad GEB is talking about Google’s new 3D cities more than they used to. I think it is quite a powerful visual effect and it deserves attention.

    Observations/questions:

    Many tree lined streams where the tree canopy covers the stream have a canyon carved through it showing the path of the stream.

    Lakes are of inconsistent quality. Some are rendered flat and others can have spikes as tall as buildings.

    Water towers are also inconsistent in quality. Many turn out better than your example, but many others do not.

    Google may manually edit bridges. Why don’t they edit coastlines? Rio de Janeiro’s doesn’t look too good. There are many other examples.

    Why do do they focus on small towns more than big cities? Also, why is coverage in some large cities so small? Chicago and London are examples.

    Sometimes I also think the rendering distance is too short. In Chicago, if you go more than 4 miles from the downtown skyline, it disappears. But in Detroit you can go much farther and still see it. It would be nice to have a setting for this.

    • Timothy Whitehead says:

      If they are using stereophotogrammetry as they have stated, and the two photos used for the stereo are not taken exactly simultaneously, then water, which is moving and lacks distinctive markings would perform particularly badly. I believe that in most cases they have manually flattened the water, but as you say there are some bits they have missed in Rio de Janeiro. Also of interest in Rio de Janeiro is the port where there are some large cranes that have almost certainly been manually edited – or rendered via a different technique, and nearby in the same port are other smaller cranes that are a complete mess. It is my guess that they have put extra effort into particularly interesting features but not bothered with everything.
      My theory on why they release small cities first is simply the amount of time it takes to get it right. London was expected for a long time, and when it was released it was clear that they had put extra effort into it. I suspect the same applies to most other large cities in the coverage areas ie they have captured the imagery but are still working on perfecting them.
      Whoever is working on it has my sympathies. It was a lot of work merely drawing a line around the edge, going through all the imagery and cleaning up particularly bad bits and fixing bridges and other attractions must take an incredible number of man hours. Maybe they should consider crowd sourcing it as they did for the various methods of producing 3D that they have used in the past.

    • As for the difference between Chicago and Detroit, Chicago was one of the first cities in the Midwest to get the new 3D. Detroit was somewhat later, and they may have improved their techniques by then.

  4. The moving object but may also explain the floating building in Fond du Lac, Wisconsin. It was under construction at the time.

  5. Wonderful posting, well explained. Has Google commented on this technology lately? They seem to be very quiet about their rollout and refinement of 3D imagery.



PLEASE NOTE: Google Earth Blog is no longer writing regular posts. As a result, we are not accepting new comments or questions about Google Earth. If you have a question, use the official Google Earth and Maps Forums or the Google Earth Community Forums.