10-05-2018 11:31 AM
Hello,
I know that it uses the Mean Euclidian Distance, but from what to what? Is it measuring the geometry of the feature polygon from the centroid to the boundary or is it some type of spectral measurement? It’s a great statistic for my project, just need to know more about the explicit details.
Thank you,
Chris
Solved! Go to Solution.
10-05-2018 12:57 PM
Hi Chris,
Presumably you are referring to the Mean Euclidean Distance Texture operator (or possibly the Mean Euclidean Distance Texture per Feature operator)?
Either way, Mean Euclidean Distance Texture is a per-raster-band moving window operation. I.e. it's measuring texture in each band. If you have an n-band image you would get n bands of texture output.
At its core, Mean Euclidean Distance Texture differences the value of each moving window cell with the center window cell value and then takes the mean of those differences. As such it's a pretty basic measure of texture and you might be better off with one of the higher order ones like Skew or Kurtosis.
The "Per Feature" variants just run the appropriate texture measure over all the pixels bounded by a feature and then take the specified statistical measure of all those values (mean, standard deviation, etc). Then that is repeated for every feature. So Mean Euclidean Distance Texture per Feature on a feature containing 100 pixels would return 100 values of Mean Euclidean Distance Texture and then calculate, say, the mean of those 100 values as the final value to attribute into the feature's attribute table. It's a way of getting a single texture value for features based on the imagery they encompass. Performing the texture calculation per feature means that you avoid introducing texture at the edges of the feature.
Does that help with the question?
Cheers
10-05-2018 01:03 PM
Fantastic!
Its for per feature over n-bands
Thank you
10-05-2018 01:10 PM
By the way - if you do want to measure the "irregulairty" of a feautre's geometry shape, there are ways of doing that in Spatial Modeler too!
Cheers