Error when writing RAT Column

[William Ray]

Hi,

I'm trying to clean up my classification by reclassifying clumps that are surrounded by a class as that class, for instance, building candidates.

I've had no problems calculating the relative border which produces a column with decimals.

However, using this new column to populate the appropriate column is showing an error:

algorithm20160214.py:160: RuntimeWarning: invalid value encountered in greater

  mod_buildings = numpy.where((rel_bor > 0.50), "building_candidate".encode(), mod_buildings)

The relevent bit of code is below:

rastergis.calcRelBorder("pleiades_segs.kea", "rel_bor", "Modern", "building_candidate", False)

rel_bor = rat.readColumn(ratDataset, "rel_bor")

mod_buildings = rat.readColumn(ratDataset, "Modern")

mod_buildings = numpy.where((rel_bor > 0.50), "building_candidate".encode(), mod_buildings)

 

I've checked using tuiview that there are clumps that match this, so that doesn't seem to be the problem.

In addition, I ran the same code specifying: 

numpy.where((mod_buildings == "building_candidate".encode()), "building_candidate".encode(), mod_buildings)

This worked fine.

I suspect it has something to do with the column format but I'm a little confused as I've had no problems with my other RAT columns.

Many Thanks

William 

[Nathan Thomas]

Hi William,

Does it populate the mod_buildings column ok, even though it gives you a RuntimeWarning?

If so, my guess is that you have a nan value in your rel_bor column which is neither equal to, less than or greater than your 0.5 value. RuntimeWarnings are fairly common in these situations although each should be double checked nevertheless.

Nathan

[William Ray]

Thanks Nathan,

I have found the solution is to name your classes using numbers. It works fine now that I have done this. I guess it doesn't like strings! 

William