[monroe74]

Originally Posted by ttoyoda

Both these methods *might* be valid if you do them on a long, dead straight road that is dead flat. Any turns or any hills up or down are going mess up the results, because you have traveled more surface miles than the "as the crow flies" miles that google/gps are showing.

You're raising two issues, curves and elevation. First let's talk about elevation.

It's true that perfectly flat roads are hard to find, but I think you're greatly overestimating the magnitude of the problem. On interstate highways, for example, I think the maximum grade is typically 6%. The average grade on the Pike's Peak race is 7%. A 6% grade would introduce a distance error of only 0.18% (derived via simple triangle math). To get an error of more than 1%, you'd need a 15% grade (or steeper). That's pretty steep. Even if you live in a hilly area, I think it should be easy to find at least a few reasonably-long road segments that don't contain any grades of nearly that magnitude.

The steepest hills in San Francisco are at a grade of about 32%. That would still introduce an error of less than 5%.

Aside from all that, Google Earth provides detailed elevation data. That means it's easy to be aware of the grade, and adjust for it. For example, I can use GE to determine that my driveway is 95 feet long, with an elevation change of 5 feet. So that's about a 5% grade.

Another tool for measuring grades and elevations is here: http://www.toporoute.com/. This tool seems to get data from maps.google. I have a feeling this is basically the same data in GE, but it's a different approach to getting at it.

OK, enough about elevation. Let's talk about curves. The GE Ruler feature has two settings, Line and Path. If you use Path, you can essentially draw curves, and therefore measure the length of a curved road. Of course the job is easier if the road is relatively straight. But it doesn't need to be perfectly straight, and it certainly doesn't need to be perfectly straight over a long distance.

The google earth ruler feature may offer a lot of digits, but I would suggest that there is no way it is good to 1/100 of an inch.

I didn't mean to suggest that it's practical to use GE to make accurate measurements down to 1/100 of an inch. Even though it reports results with that level of precision. However, for our purpose, we don't need accuracy down to 1/100 of an inch. We just need a few feet, and GE can definitely do that. For example, it has no trouble telling me that the fire hydrant in front of my house is 22 feet from my driveway. Or that the skylights on my roof are 10 feet apart. I know that these results are correct, within a few inches. If instead of using landmarks 10 feet apart, I use landmarks a mile apart (and GE obviously lets me do that), then I can measure a mile with an accuracy of a few inches. This is very competitive with any other method I know of.

You cannot even resolve where the sloped edge of a curb, or the edge of a road starts from those views.

It could be that this varies depending on what part of the world you're looking at. If you look at Times Square, for example, you can zoom in so your eye altitude is about 200 feet, and then it's pretty easy to find out that 45th St is about 33 feet wide, curb-to-curb. Do I have this answer to 1/100th of an inch? Of course not. But I don't need that.

So for the roads I'm looking at, GE definitely lets you clearly see where the curb ends and the road begins. But it could be that for certain areas, GE is using photos with lower resolution, and you might be looking at one of those areas.

For comparison, let's consider the traditional approach to measuring distance: your odometer. We all know that changing tire size will introduce an error. But what about tire wear? If you do some calculations, you realize that worn tires will produce an error of about 1% in your odometer reading. In other words, the error introduced by normal tire wear can very easily exceed the error introduced by using GE and not bothering to account for grades. And it's easy to account for grades and therefore eliminate even that small error.

What is the accuracy of GPS now, +/- 30 feet?

Civilian GPS is accurate to about 50 feet. So you're right, GPS is probably not the best way to do a VSS distance calibration. But it's still probably better than the traditional method: your odometer. Unless you have a better idea, but you haven't mentioned what it is.

To summarize, these are the most obvious methods I know about, to calibrate automotive distance, in descending order of what I think the accuracy is:

A) Google Earth
B) Mile markers placed by your highway department (as mentioned by diesel john)
C) GPS
D) your odometer

Regarding D, it should be noted that they are not famous for being accurate, even before you start worrying about things like tire size and wear.

dcb, sorry for the tangent. I know you want to keep this thread focused on development issues, but this seems at least somewhat relevant. Hopefully.