Re: An interesting problem: distance estimating

["markw-4" <user=5@nabble.blu.org>]

 > Matthew Gillen wrote: 
>> [hidden email] wrote: 
>> 
>>>> On Thu, Jan 03, 2008 at 10:16:09AM -0500, [hidden email] wrote: 
>>>> 
> ... 
>>>> So adjust to suit your intuition: 
>>>> 
>>>> - for x > 250 miles, return x 
>>>> - for 100 < x < 250 miles, return 1.05 * x 
>>>> - for 10 < x < 100 miles, return 1.1 * x 
>>>> - for x < 10 miles, return 1.3 * x 
>>>> 
>>> That just seems too arbitrary, there should be some curve I could use 
>>> :-) 
>>> 
>> 
>> You could use google maps' web service api to experimentally gather some 
>> data 
>> points, then use your favorite data analysis package to fit a good curve 
>> ;-) 
>> 
>> Matt 
>> 
> I would say use Google Maps to determine the fractal dimension of the 
> roads, then use that dimension in your distance calculations. 
> That dimension may itself actually vary depending on the "flying crow" 
> distance, but would probably give you a fairly accurate measurement over 
> all. 
> 
> That's assuming, of course, roads are scale-invariant, and I'm not too 
> sure of that. Well, some research for you to do!