My Google Map Blog

A recent story in the news says that Taiwan’s Defense Ministry has asked Google to blur imagery of some military structures on Itu Aba Island (also known as Taiping Island) in the South China Sea.

The South China Sea is a hotly contested area with several countries, most notabley China, building structures on the reefs of the Spratleys. We had a look at the progress last year and had a look at the latest imagery of Fiery Cross Reef in June this year.

As we have discussed before, censoring satellite imagery is more difficult than censoring other data. A number of countries censor aerial data, mapping data and 3D imagery and because those are often gathered within the countries in question, Google must comply with local laws. But satellite imagery is more difficult to censor. I has been done. It is illegal in the US to publish high resolution imagery of Israel, and Google is a US company so Israel appears in relatively low resolution in Google Maps and Google Earth. Also, imagery updates were stopped for Iraq, Afghanistan, the Ukraine and Syria although we do not know exactly how that was accomplished.

Back to this particular case, if the story is correct and if that was the only facility that Taiwan asked Google to censor, then it would appear they made a serious error. The result has been that many news outlets have published the imagery and it is now not only practically impossible to get rid of all those images, but it has become an object of attention. In addition, Google will be very unlikely to agree to censor it unless legal pressure can be brought to bear. If they were to censor it, a lot of people will notice, and it will become another news story, and then a lot of other countries will want their censorship requests to be honoured as well.

Another question would be who Taiwan wishes to hide the imagery from. Even if Google removed the image, it could still be purchased from the supplier Digital Globe, or imagery of the location could be ordered from various other commercial suppliers. A number of states also have their own spy satellites and the countries most interested in the region could probably capture aerial imagery of it if they wished.

The image that Taiwan wants censored is dated July 8th, 2016 and can only be seen in the default layer as it has not yet (as of this writing) been put in historical imagery.

The military structure in question can be seen below:

We have also created an animation showing that the airport was built since 2006 and the harbour is also being expanded.

Speed in milliseconds per image:

The post Taiwan wants Google to censor satellite imagery appeared first on Google Earth Blog.

The Erskine Fire was a wildfire near Lake Isabella, California. It was only one of many wild-fires that occur in the California region every year (there are an average of around 8,000 fires per year). According to Wikipedia, the Erskine fire cost US$ 19.3 million, destroyed 309 buildings and caused two fatalities.

We noticed a reddish brown strip that appears to largely match the edge of the region that was affected by fire. We cannot tell whether it is a natural geological feature (that for some reason stalled the fire), a result of the fire itself or a result of fire-fighting efforts. If any of our readers knows what it is, please let us know in the comments.

From the aerial imagery the region appears to have very little vegetation, but it appears to have been more than enough to create an uncontrollable inferno.

Here are some ‘before and afters’ of the damage caused by the fire:

This was the worst affected area, a suburb of a small town called South Lake.

This KML file has placemarks showing the locations of the damaged buildings we were able to find as well as an image overlay from the map shown on Wikipedia showing the approximate extent of the fire.

The post Damage by the Erskine Fire, California as seen in Google Earth appeared first on Google Earth Blog.

Artificial intelligence based facial recognition is improving over time. However, it is a bit too good at recognising faces as two recent stories illustrate. The first problem is Google’s automatic facial recognition as used to blur faces in Street View imagery, tends to err on the over eager side and ends up blurring faces of statues, people in paintings, and now even a cow. Read more about it here (expect a lot of bovine puns).

A cow in Cambridge, England, has its face blurred for privacy reasons. See in Street View

The second story found here and here says that artificial intelligence based facial recognition can still identify faces fairly accurately even with face blurring. This would suggest that Google’s efforts to blur people’s faces in Street View may soon be thwarted and they will need to redo it all with a more secure method. The easiest reliable method is to completely cover faces with a square of solid colour. Another sightlier alternative would be to subtly warp faces in addition to blurring them so as to fool facial recognition algorithms. This should work until reliable body recognition becomes common place (yes that’s a real thing and probably works even better on cows than facial recognition does).

The post AI facial recognition is too good, in two different ways! appeared first on Google Earth Blog.

Although Google has neglected to update ‘historical imagery’ in Google Earth since early June, they have been adding fresh imagery, and when it is reasonably good quality, it goes into the default layer and we can see it. One such instance is a region in Jiangsu Province, China which was struck by a deadly tornado on June 23, 2016. According to Wikipedia, the tornado killed at least 99 people and injured 846 others (152 critically).

We found a number of articles showing various photos of the destruction, such as here, here, here and here. But, actually locating the event proved more difficult. We first mapped out the area that had new imagery and started searching through it for signs of damaged buildings, but with an area of nearly 4,000 square kilometres we were not successful. We did find a raised railway under construction, and a long trail of destroyed houses that turned out to be planned road construction. The articles either mention major nearby cities or small villages that aren’t marked on the map and couldn’t be found through search. Eventually we found mention of “Danping Village of Chenliang Township” and we were able to find Chenliang. From there, the path of destruction was easy to trace over a distance of around 30 km.

To see the path of the tornado in Google Earth download this KML file.

Although the latest imagery is not very high quality, near total destruction of houses all along the path of the tornado is clear, especially when comparing it with older imagery.

Some damaged factories.

A factory roof ripped to shreds.

See a higher resolution aerial image of this factory in this article

The post The Jiangsu tornado appeared first on Google Earth Blog.