current methods of augmented reality - geo tags vs. trackables (part1)

hi,
i want to talk about two basic methods that can be used to create augmented reality apps on mobile devices. today i will write about the first one. the other one will be covered in another post soon.
let's take it as given that the current method to generate augmented reality is to use the camera view and  generate an overlay. this is usually done by using tags to present additional information or by calculating the correlating view of a 3d model.
the first method is already in use since quite a while. i'm talking about geotagging. the basic idea behind is that 2d tags or 3d models are stored in a database with their according geographical location information (gps coordinates, hight above ground). to view the tags and models in their correct location on the screen for the current camera view some values have to be measured from three sensors:

• gps: to get the actual position of the device; user can set a range around this location in which models or tags are shown.
• compass: to get the horizontal orientation of the device
• accelerometer: to calculate the vertical orientation of the device.

the combination of these values should provide the application with all necessary data to display the elements in the correct position.
while this method seems to work quite well in general (similar solutions can be found in apps like layar or wikitude), there are some serious problems. they are actually all related to accuracy and responsiveness and are caused by two facts. the first one is the general problem when working with gps. as long as you are in a rather open area the positioning works pretty well, but as soon as you are in a city with high buildings accuracy and speed of repositioning seems to decrease rapidly. Another problem is caused by the fact that data from three different sensors is used. this gives a lot of possibilities for mistakes or responsiveness issues and might affect the final result drastically. one more issue related to 3d models is, that occlusion by real world objects can't be taken into account so far. if the radius to display models is set too high, you can easily end up seeing a 3d model which shouldn't be visible (because it should be covered by a real building). so this technology works best with a very small radius and a low density of 3d models at once.
finally, as gps only works outdoor properly, this method can't be used inside a building unless another technology is used for the positioning.

i hope this gives you some idea about this method. the second part can be found at the link below.
current methods of augmented reality (part2)