[asxless]

What is the accuracy of the GeoTags attached to photos taken with the iPhone Camera app on a iPhone 3Gs?

Note: I'm not asking about the 'precision' of the GeoTags as implied by the number of digits. I'm asking about the 'accuracy' of the GeoTags. IOW in an open field, with a clear view of the southern sky, how accurate are the GeoTag locations relative to one another?

I ask because my first attempt to use the Camera app GeoTags was spectacularly disappointing. I took nearly 200 photos while walking in a area roughly 600 ft N-S and 1,800 ft E-W. ALL of the GeoTags fell on a .6 second square latitude and longitude grid. Not only was every GeoTag either dd mm s.0 or dd mm s.2 or dd mm s.4 or dd mm s.6 or dd mm s.8; Adjacent GeoTags were at least .6 seconds apart. IOW the GeoTags implied that I had somehow managed to take every single picture precisely .6 seconds north-south and/or east-west of nearby photos. For the idly curious, grid points in a .6 second Lat/Long grid are roughly 50 ft apart. This implies that the GeoTags accuracy is +/- 50ft. BUT....

The accuracy of the iPhone GPS appears to be much better than +/- 50ft. implied by the Camera App GeoTags. In fact, during this same photo shot I switched to the Google Earth app several times to compare the little blue dot to know locations on the ground imported via Google Maps. Each time I was very impressed by the accuracy of this display. Which begs the question -- why are the Camera app GeoTags so crude?

asxless in iLand

[mduell]

The horizontal precision of civilian GPS with a single receiver is about 50 feet. Complex set ups with multiple receivers or multipatch antennas can do better in other devices.

[Cold Warrior]

Sometimes I've seen pics not get geotagged at all or well. A tip that works for me is to open up Maps and hit the gps button. Let it lock in on a good location with gps and possibly a-gps, then return to Camera. Subsequent pics will have improved geotags.

When the gps fix is good, I see 50 ft or less precision.

If you were to look at the embedded exif gps information, you might see the level of precision. For example, I've seen it out into stuff like latitude xx"xx'xx.59999 longitude xx"xx'xx.799995

[asxless]

Originally Posted by mduell 

The horizontal precision of civilian GPS with a single receiver is about 50 feet. Complex set ups with multiple receivers or multipatch antennas can do better in other devices.

HUH? Maybe you are talking about ABSOLUTE Lat long accuracy. But I am talking about relative accuracy over a few hours.

FWIW I regularly get much better than 50ft GPS relative accuracy with my Garmin when tracking or taking way points over a few hour period. And even the iPhone produces better accuracy than 50ft as visualized in Goggle Maps and Google earth during a walk in a open field. For example, it is fairly easy to walk up and down a jeep trail and be able to tell which rut I walked up vs which one I walked back. IOW the relative accuracy of my Garmin is better than 15ft. And in limited testing the iPhone is about the same. If I repeat the same walk on another day, the tracks will probably not exactly overlay BUT I will still be able to distinguish which rut I walked up vs down And more importantly, way points taken along the walk will not 'precisely" fall on grid points .6 seconds (roughly 50ft) apart).

AFAIK the issue is NOT the accuracy of the iPhone's GPS. It is the accuracy of the GeoTags recorded by the Camera app.

asxless in iLand

[ghporter]

Relative accuracy, minute to minute, depends on multiple factors. Since EACH fix can be easily 50ft or more "off" the absolute location it is taken at, you can have two fixes just moments apart that are many times that 50' apart. GPS almost always generates each fix uniquely, without reference to a previous fix. As you're standing there, one particularly clearly visible satellite could set while two more crappy ones (as far as your location is concerned) could rise, producing a drastic change in the absolute accuracy of subsequent fixes. Further, you could be out in the open, but with the "best" satellites (those near your horizon) hidden by buildings and trees, which would leave you with data from relatively poorly located satellites that are nearly directly overhead.

It's not like a GPS unit reads hidden lines on the ground or something; a GPS receiver has to pick out signals that are intentionally lower than background noise, sync itself with them, decode the data, then perform a complex matrix algebra computation. Many times a second. If the RF signal, broadcast at (at most) 50W by each satellite in its orbit of over 11,000 statute miles in altitude, is poor due to atmospheric effects, obstructions, or even sunspots, it doesn't matter how robust the processor and decoder in the receiver are-with garbage RF in, you get crappy GPS data out.

[Wiskedjak]

Wow ... I thought the military had given up on restricting civilian GPS accuracy over a decade ago since it's easy enough to get accurate positions using multiple receivers.

[asxless]

Originally Posted by ghporter 

Relative accuracy, minute to minute, depends on multiple factors. Since EACH fix can be easily 50ft or more "off" the absolute location it is taken at, you can have two fixes just moments apart that are many times that 50' apart. GPS almost always generates each fix uniquely, without reference to a previous fix. As you're standing there, one particularly clearly visible satellite could set while two more crappy ones (as far as your location is concerned) could rise, producing a drastic change in the absolute accuracy of subsequent fixes. Further, you could be out in the open, but with the "best" satellites (those near your horizon) hidden by buildings and trees, which would leave you with data from relatively poorly located satellites that are nearly directly overhead.

It's not like a GPS unit reads hidden lines on the ground or something; a GPS receiver has to pick out signals that are intentionally lower than background noise, sync itself with them, decode the data, then perform a complex matrix algebra computation. Many times a second. If the RF signal, broadcast at (at most) 50W by each satellite in its orbit of over 11,000 statute miles in altitude, is poor due to atmospheric effects, obstructions, or even sunspots, it doesn't matter how robust the processor and decoder in the receiver are-with garbage RF in, you get crappy GPS data out.

I appreciate everyone's input BUT... There seems to be a severe gap between the theory being described and my actual results with hand held consumer grade GPS units.

FWIW I have used two hand held Garmin consumer grade GPS units at lot over the last half decade. I regularly get way points at know locations (e.g. property corners, etc.) that are repeatable from year to year, season to season, day to day and hour to hour that are better than 15ft in ABSOLUTE accuracy. By that I mean that I can walk out tomorrow and take a GPS way point at a property corner (i.e. an iron rod in the ground) and it probably be within +/-15 ft of way points I have taken at that same location at different times over the last 5 years. In fact in any given GPS session, a series of way points at several known locations are usually much better than +/- 15ft among themselves. IOW the way points taken yesterday and the ones taken a few weeks ago may differ by +/- 15 ft BUT the two sets of way points can be 'adjusted' by simply translating one set relative to the other to reduce this error. Hence the relative error within each set is LESS than +/- 15 ft.

If consumer grade GPS units were as bad you and others have described they would not be able to be used in car navigation systems or to track hikes, etc. 50 ft is wider than most 2 lanes roads, and certainly wider than most walking trails. Yet hand held GPS units regularly track cars and people accurately enough to show that they are in fact on the road or the trail The GPS built into the iPhone does this pretty well even in sub optimum signal conditions. During the photo shoot I described in the OP, I tracked my path walking a property line using the Google earth app on the iPhone in moderately heavy woods. I was impressed with how well the little blue dot tracked the know property line that I had uploaded via Google maps. BTW I've walked this same property line at least a dozen times using a Garmin hand held. The results were quite comparable.

asxless in iLand

[ghporter]

Dedicated GPS devices tend to take FAR more samples and generate their updates many more times a second, so they should be more accurate and self-consistent. Further, your Garmin probably has multiple, unique receivers (my GPS-12 has (you guessed it) 12 separate receivers), so it can "see" multiple satellites at the same time, while a single receiver, or even a 4-channel unit can only track 1 or 4 satellites at a time, thus reducing the precision of its output.

And my post was not theory. That is the way GPS works.

[KeyLimePi]

I don't know anything about GPS, but I know my gf took some pics of her garden with her iPhone. I opened one in 'Places' in iPhoto, and when I used the satellite view it put the pin in the exact spot where our garden is in our yard. For what most people use it for, I think that it's pretty decent.

[Dork.]

Is there a spec for what information is in those GeoTags? It sounds to me like the full position information (to the precision that the iPhone's GPS generates it) is not being reported properly. Regardless of the precision that the Camera app reports to you, it appears that the actual information the Camera app has is not as precise. To me, this either means that the GeoTag does not have enough information, or the Camera app cannot resolve the information it has to less than 50 ft.

(I don't own an iPhone and I don't know anything about the Camera app or GeoTags, this is all speculation based on the OP).

GPS accuracy depends on the number of satellites used in the fix, and how far apart they are. Given a clear view of the Southern sky (assuming you're in the Northern hemisphere, of course), you should not be having problems. Isn't there an app that tells you about the quality of the GPS fix and where the satellites are? I found one for my Droid, there must be one for the iPhone. I seem to recall Trimble having an iPhone app, but I don't think it's free....

[ghporter]

It could also be that Camera.app limits the accuracy of the data used in the tags for some reason.

Oh, and you don't need "southern sky" exposure; the GPS satellites are in a complex set of orbits that cover the entire planet. Only the poles have "limited" coverage, and even those areas are covered by at least 4 satellites at all times.

[Dork.]

The GPS satellites orbit between 55 Deg. N and 55 Deg. S. So, strictly speaking, you can see satellites overhead or in any direction as long as you are in that band. And you do only need 4 satellites for a fix.

But your position fix gets better based on how many more satellites are in view, and how far apart they are. If your Southern exposure is blocked, you may have trouble getting enough extra satellites (that are spread out enough) for a "better" fix.

I'm not familiar with the chip in the iPhone, but most consumer level receivers can use 12 or more satellites in the fix.

[asxless]

For those who prefer doing physical experiments to speculating....

Go to an open field and lay out a line approx. 150-160' long at roughly 320 degrees (NW-SE)*. A 150+ foot measuring tape would be handy. Then take your trusty iPhone 3Gs out. Give it a little time to get a good fix on available sats; you can estimate this by the size of the circle in the Google Map app. Then begin walking the line and taking a photo (of the measuring tape) every 10ft. When you are done, sync your iPhone with iPhoto. You can view the tags directly in iPhoto. But it helps to use the free OS X application iPhotoToGoogleEarth iCraig � iPhotoToGoogleEarth to visualize the geotags in Google Earth.

You will most likely end up with 4 clusters of geotags on a square grid separated by 'precisely' .6 seconds in both Lat. and Long. If you are lucky (or cheated) two of these clusters will fall directly on your walking line. But more likely, two clusters will be SW and two will be NE of your line.

To check your results you can use a hand held GPS receiver (e.g. a Garmin Vista, or equivalent). Turn on tracking and take waypoints every time you take a photo. Upload these to Google earth for a little insight into the relative accuracy of the iPhone's Camera apps geotags.

And for extra credit you can use the MotionX iPhone app (or equivalent) to compare to the hand held GPS receiver's track and way points and the iPhone Camera app's geotags.

The comparisons between the three data sets may surprise you

asxless in iLand

* 320 degrees is roughly the bearing of a diagonal between points in a .6 second lat/long grid at roughly 30 degrees latitude.

[Dork.]

.6 seconds is the increment you get if you divide a minute into 100 divisions. More speculation, but I'll bet the GeoTag is not even storing seconds at all, but just storing position down to .01 of a minute. Totally accurate, but lacking in precision.

[asxless]

Originally Posted by Dork. 

.6 seconds is the increment you get if you divide a minute into 100 divisions. More speculation, but I'll bet the GeoTag is not even storing seconds at all, but just storing position down to .01 of a minute. ...

Bingo!

Based on my experimentation,
* The Camera app on an iPhone 3Gs rounds (or possibly truncates) the photo geotags to .01 minutes. This forces all the geotags onto a .01 minute (.6 second) grid
* iPhone GPS apps like MotionX do quite a bit better. At least all the MotionX waypoints fall on a track and not on a grid

But the bottom line is that I won't be retiring my Garmin eTrex Vista just yet

asxless in iLand

p.s. the reason I initially reported the iPhone geotag grid as .6 seconds is because iPhoto displays the geotags in dd mm ss.sss. So anyone inspecting their own geotagged photos in iPhoto could see the pattern more easily.