3d imagery

We recently came across the YouTube video below about Google Earth’s 3D imagery. It gives some insight into how it is gathered and we also get to see some of the faces behind Google Earth, such as Google Earth Product Manager Gopal Shah.

Apparently the aircraft used to gather the 3D imagery use five cameras, one facing down then four others pointed in different directions. The ‘stereo’ imagery is not actually achieved by two cameras taking photos from different angles, but rather each single camera taking multiple photos as the aircraft is moving, effectively achieving a stereo effect. Previous analysis we have done (1 2 ) suggests that each camera captures four images in quick succession to get the ‘stereo’ effect for any given location and then imagery from multiple passes from different sides of the location are combined to create the final 3D.

Also interesting is that Gopal seems to imply that cars are manually removed from the imagery rather than via an automated process. Not mentioned in the video is the fact that Google pays special attention to some structures like bridges, harbour cranes and certain intricate buildings, using more manual methods. We believe the manual part of the process is the main reason 3D imagery often takes a long time between when the images were captured and when they are published in Google Earth.

It must also be pointed out that Google is continually improving the process and the latest releases of 3D imagery are significantly better quality than the oldest 3D imagery.

Google has recently released a new batch of 3D imagery. The last major release was in early February. Thank you to all the GEB readers that contribute by finding new areas and drawing outlines. Below is a chart showing the areas in square kilometres released over time. It does not include all of the recent March batch as only some of the outlines have been drawn so far.

To see the area covered by 3D in Google Earth download our KML map.

Keep in mind that the above areas do not include updated regions. These can be difficult to identify so we do not track them. The general trend is a decline in new area covered. We believe that this is partly a reflection of the fact that most of the large population centres in countries where Google is releasing 3D imagery have already been covered. It is likely that Google still releases approximately the same area of 3D imagery each month, but much of it is now updates, not new areas.

The largest single area at 19,500 square kilometres is now the New York City region. It includes parts of five different states and is more than double the area of the second largest region (San Francisco at 7,800 sq km).

Thank you to GEB reader Ryan for pointing us to Trump’s aircraft captured in 3D in one of the latest 3D releases in Florida:

It is parked not far from Mar-a-Largo, owned by Trump and which he now refers to as his Southern White House. We were unable to date the 3D imagery as there is not much aerial imagery in that location to compare it with. Trump now uses Airforce One, which we have also looked at in 3D. Find it in Google Earth with this KML file.

If you ever get bored of Google Earth’s 3D imagery, then a fun thing to try is changing the ‘terrain exaggeration’. It is a setting found in Tools->Options->3D View->Terrain (PC) or Google Earth->Preferences->3D View->Terrain (Mac). This setting stretches or compresses the terrain vertically, and it also applies to Google Earth’s 3D imagery. It can give a whole new look to familiar sights.

Table Mountain, Cape Town, South Africa. No terrain exaggeration.

Table Mountain, Cape Town, South Africa. Terrain exaggeration: 3.

Cape Town, South Africa. No terrain exaggeration.

Cape Town, South Africa. Terrain exaggeration: 0.3.

New York, United States. No terrain exaggeration.

New York, United States. Terrain exaggeration: 3.

Remember to set the terrain exaggeration back to 1 once you are done.

While exploring Google Earth’s 3D imagery, you will probably have noticed that even in the busiest cities, people are largely absent. This is because of the way the 3D imagery is captured. Based on previous analysis, the 3D imagery is captured by taking four images in rapid succession from an aircraft and then another four later on on a separate pass. It is also possible that more than two passes are combined, and we have even come across one case where it appears that Google combined imagery from completely different flights months apart. The overall result is that anything that is moving cannot be properly imaged in 3D. Over time, Google has got better at removing moving vehicles from the imagery so the streets appear almost empty, and we never seem to see pedestrians at all.

However, while exploring the 3D imagery in Dublin, Ireland we came across some people in the parks there. It mostly includes people lying down or sitting and they evidently stayed in the same place long enough that the 3D imagery process was able to capture them. However, as far as we can tell, none of them actually got proper 3D models. That is probably due to a combination of factors. The resolution of the 3D mesh is not really high enough to model a human. Also if the people were moving around even slightly, although they will still be visible in the image data, the algorithms used to generate the 3D from stereo images would fail.

Dubh Linn Gardens, Dublin, Ireland

We also managed to find some people on a beach in France. In this case a few of the beach umbrellas did get detected by the 3D generation algorithms. The waves on the shore seem to have caused problems for the algorithms as there are blobs of floating sand all along the shoreline.

Beach in Marseille, France

Beach in Marseille, France

Have any of our readers come across a place where a person was standing still long enough to get captured in 3D? Let us know in the comments.

Thank you to GEB reader Jacob for bringing to our attention an interesting effect in Google Earth’s 3D imagery. In Kingston, Ontario, Canada, there is a building in the 3D imagery that seems to be a combination of imagery from when it was still under construction and imagery from after its completion. The result is quite interesting and worth exploring in Google Earth. Find it in Google Earth with this KML file. Be sure to turn on the 3D buildings layer.

The 3D imagery was only recently added to Google Earth (first reported by GEB readers on January 3rd, 2017) but, based on historical imagery the building was constructed between May 2014, when there is no building visible, and September 2015 when it is mostly finished. So at least some of the imagery used to create the 3D is several years old.

We have previously noted a case where a building could be seen in different stages of construction depending on zoom level and an instance where whole buildings were missing from the 3D imagery.