I have been exploring the GPS data mining literature esp. for problems like anomalous trajectory detection, time travel prediction, etc and one very common method I see is dividing the data or map into grids. Can any one please explain the logic of this? Are the coordinates euclidean in this case? Is grid decomposition really necessary?
I would be grateful if someone can also give/ quote some links or materials I should explore. I am new to this field, so please pardon me if the question is very obvious.
No they are not euclidean. But they don't have to be. The grids are not rectangles anymore, but can be treated as such for some operations.
If you create a lat/long grid, then each cell by means of meters is not rectangular. However it defines a zone where you add a counter, which has a clear inside/outside definition. And you can use cartesian operations (Rectangle.inside())
So the lat / lon span is constant for each cell, but not the longitudinal meters span, which shrinks by cos(latitude).
If one needs a grid with equal grid cells sizes by means of meters, then one
has to transform the geo coordinates before.
Asked in February 2016Viewed 2,674 timesVoted 4Answered 1 times