[Geowanking] Cheap post-process DGPS - why not?

[stephen white]

On 06/08/2007, at 11:13 AM, Ben Discoe wrote:
> 5. Using the timecode to correlate, subtract the second unit's  
> drift from the first unit's coordinates.

I can't see how you would do this, because you wouldn't be able to  
tell the difference between movement and drift. The second unit would  
be drifting around at random, rather than having a fixed offset that  
you can add or subtract from the first unit.

> But it should be a really simple operation to subtract one track's  
> offset from another.  Is there some reason this simple approach  
> wouldn't work?  Is there some FOSS which will do it?

I would need to know the situation where you need that level of  
accuracy before I could recommend a range of possible solutions.  
There are basically no solutions that let you walk around with  
accurate tracking no matter how much you spend. You need to pick the  
solution based on the strong points, then wrap the rest of the  
package around fitting to those points.

For example, I use orthorectified aerial maps along with photographs  
to get about 5cm accuracy at selected points. That technique doesn't  
work indoors (obviously) so I have to shift to other methods like  
camera calibration and triangulation to extrapolate a series of  
photographs through scenes. This too would not work for extended  
scenarios due to accumulative error.

The combination of the two inferior methods achieves what I want as  
the aerial imagery provides the broad scale accuracy, limiting the  
indoors extrapolations to short stretches from the front door.

--
   steve at adam.com.au