Geotagging and Augmented Reality – New Standards Needed?

Augmented reality (AR) apps are all the buzz at the moment, and they do offer some exciting possibilities. In fact I have already created an iPhone app using certain AR techniques myself, and intend to submit it to the app store as soon as Apple will permit it. More later. 😉

However, whilst considering some of the implications of using AR, I was surprised to find that current geotagging standards don’t seem to be on a par with what AR technology permits. There is definitely scope for extending the current standards, and its seems very likely to me that demand for this will grow rapidly from now on.

In its simplest form, geotagging is simply a means of tagging a piece of data (such as a photo) with a latitude and longitude, and thus allowing data to be mapped or accessed via locational search.

An implicit assumption (given that it is called geotagging) is that the data is associated with a point on the earth’s surface, whatever the altitude of the surface might be at that location, in which case an altitude is not strictly required, because the earth’s topography can be assumed to be reasonably constant and knowable via other sources. But what about the case, for example, when a photo is taken from an aeroplane? In that case, altitude would be an additional parameter required in order to correctly differentiate it from a photo taken on the earth’s surface. Given that GPS systems typically record altitude, it is hardly a stretch to include it as  standard additional tag item.

And going further still, given that augmented reality devices measure spatial attitude of the camera, would it not make sense to enable any photo taken to be tagged with the heading (aka azimuth) and elevation (ie. local horizon coordinates) at which is was taken, and further also with the camera tilt (eg. whether portrait or landscape relative to horizon, or any angle in between), and with angle subtended by the camera shot, both vertically and horizontally. It would be necessary to provide all this extra geotagging information in order to be able to correctly reproduce/model the exact “window” into space that the photo represents. Whilst this additional data might would be unnecessary for common scenic or portrait photography, it could be quite valuable in other situations. For example, two such fully (and accurately) geotagged photos taken of the same object from different locations would allow a 3-D reconstruction of that object to be created. This is not a trivial implication!

One of the existing geotagging standards (FlickrFly) also allows an additional item to be specified ie. distance or range from the camera to the subject of the photo, which presumably is necessary when a locational search is being done for the subject of the photo rather than for the location from which the photo was taken.

In order to avoid missing out on the possibilities that AR apps can already provide, and to be able to start constructing better, more powerful geolocational systems and applications which take advantage of this, I would propose that existing geotagging standards be extended to include all of the following:

  • date, time, timezone offset (to provide instant in time)
  • geographic latitude, longitude and altitude (to provide location relative to earth’s surface and mean sea level)
  • camera viewpoint central heading (0°=N, 90°=E, 180°=S, 270°=W)
  • camera viewpoint central elevation (0°=horizontal, -90°=vertical downwards, +90°=vertical upwards)
  • camera tilt (0°=portrait/upright, 90°=landscape/top points left, 180°=landscape/top points down, 270°=portrait/top points down)
  • camera angle subtended vertically (ie. along nominal top to bottom)
  • camera angle subtended horizontally (ie. along nominal side to side)
  • range of point of interest from camera (if there is a relevant POI involved)

I would welcome feedback on this proposal from anyone knowledgeable with the current state of geotagging. I am certainly not an expert in this area, but for those who have the capabilities to influence things in this area, I believe that there may be opportunities for valuable advances, especially in relation to the new generation of AR technology.

Estimated Total Value of App Store Market

AdMob has just issued a report with some very interesting data on the app store, and one of their claims is that the app store market is worth about US$200M per month.

This is disputed as being an unreasonably high estimate by some commentators. However, with data I have gathered from my own app sales, I am in a position to make my own estimate too. So what is it? Well there are four steps.

Step 1 – Sales Versus Rankings

In an earlier post I showed a graph of sales versus rankings for Oz Weather in the Australian app store during the first few months of 2009. Below is an updated graph covering the sales period July/August 2009.

Oz Weather Paid Rankings

Although Oz Weather hasn’t ranked highly enough to provide data points within the top 10, it does still allow a good estimate of the curve as a whole ie.

Daily Sales = 1800 * Ranking ^ -0.8

Note that this curve is substantially higher than the 6 month-old estimate, showing a large increase in the overall number of app sales per day – of the order of a factor of  2 or more.

Step 2 – Find area under the curve to give total daily app sales

The area under this curve is the integral of the curve formula. With some high-school calculus, we might know that this is:

Area under curve = 1800 * (Ranking^0.2 / 0.2)

Solving this for the range 0 to 50,000 apps (the approximate number of paid apps available) gives the result:

52,000 app sales per day in the Australian app store

Step 3 – Multiply by average app price

Assuming an average app price of US$1.80 (AUD$2.25) the total daily revenue for the Australian app store alone is

  • AUD$120k per day
  • AUD$3.6M per month
  • AUD $42.7M per year

Step 4 – Expand to whole Globe

I have previously estimated the Australian market to be about 1/30th of the global market size. However, the AdMob report indicates about 45M iPhone/iPod Touch devices worldwide, and previous research has implied about 1.1 million in Australia, implying a ratio of about 1/40. In that case, the corresponding global sales figures would be

  • US$3.9M per day
  • US$115M per month
  • US$1,370M per year


Although there are a number of estimates and assumptions involved in these calculations, the final monthly number of US$115M per month, although less than AdMob’s estimate, only differs by a factor of a 0.6.

A more accurate estimate would be possible with a wider range of app versus ranking data – especially from the larger app stores like the US. However, I did previously collate some data like this, and it also supported a rankings/sales tail off with a power factor close to -0.8.

I conclude that my own app’s sales figures imply that AdMob’s estimates are very plausible, and certainly likely to be in the correct ballpark.

Oz Weather Apponomics – Part 7

This is the latest installment tracking the progress of the Oz Weather iPhone app in the iTunes app store. (Part 6 installment here.)

The latest stats, to 4th August 2009:

* Total app downloads: 48,600
* Net app revenue: AUD$73,800 (US$59,000) – net of 30% Apple share and 10% Australian GST
* Average User Rating: 4 stars – from 885 ratings of all versions

The following graph shows a complete history of more than 8 months of daily sales records, since launch on Nov 1st 2008.


The associated Australian overall paid apps ranking is as follows:


I already explained the cause of the great sales dip in June in the previous post, but the other feature that stands out here is how the ranking in the latter half of the graph has been declining despite a fairly constant average base level of sales (excepting June). The obvious explanation for this divergence is that the total number of all apps being bought is gradually increasing with time as the number of iPhones/iPods in Australia has increased. Recent estimates by AdMob put the total number of iPhones in Australia at around 750,000 and iPod Touch at 350,000 – making a combined total of 1.1 million devices on which Oz Weather could be installed. This would mean that Oz Weather has been purchased by about 4.4% of Australian device owners.

This might seem to leave room for plenty more sales, but others have suggested that 3% is a high ownership rate for other popular apps such as “Flight Control”, so maybe we’re already pushing the boundaries!