Finding routes

Discussion created by modybsystematics-co-il-esridist Employee on Aug 20, 2012
Latest reply on Aug 21, 2012 by modybsystematics-co-il-esridist
Hi all

I need some ideas on how to perform some analysis. Here is the problem.
We have some GPS points of moving cars.
The GPS is not accurate and the delta from the real location can be 50 meters or more.
We create lines from the points. We might snap it to roads network but in some cases in city the points can snap to the wrong street.
What we are looking for is places where the person drove only once (or only twice).
Because the route is not accurate we must translate the lines into raster in some kind of Kernel Density process and then check the raster.
The problem is that if we have two lines the raster value is higher than if we have one line but when we move away from the two lines value we get similar values to the one line values.

Can anybody think about a nice way to do it?