Cloud mask for worldview

Question

I'm using worldview imagery downloaded from gbdx and I want to apply cloud mask for it. It's preferable to use ready-made layer or some python module/tool to create it. The demand for the method precision: it's allowable to cut clouded areas with some reserve (lose a part of useful information).

What should I try to consider?

I thought about some threshold, but perhaps it would be a bad idea.

Answer 1

This report uses the Forest Discrimination Index (FDI), described by Bunting and Lucas (2006) to create a mask. Essentially, the FDI for worldview2 is NIR2 - (RE + Blue), or band8 - (band6 + band2). You have to play around with the threshold, but here is a function I used in arcGIS Pro. One of the problems is that it will cut out buildings, roads, etc. For my purposes, I don't care. Others may not want this.

def CloudMask_WV_FDI(infile,threshold,outfile):
    """ this function attempts to mask clouds using the 
    forest discrimination index (FDI), as outlined in Bunting and
    Lucas (2006). """
    band8 = infile + "\\Band_8"
    band6 = infile + "\\Band_6"
    band2 = infile + "\\Band_2"
    outraster = arcpy.Raster(band8) - (arcpy.Raster(band6) + arcpy.Raster(band2))
    outraster2 = arcpy.sa.FocalStatistics(arcpy.Raster(outraster), "Rectangle 10 10 CELL", "MEAN", "DATA", 90)
    outraster2.save(r"FDI")
    del outraster,outraster2

    outraster = arcpy.sa.SetNull("FDI", 1, "VALUE <= -0.2")
    outraster2 = arcpy.sa.ExtractByMask(infile, outraster)
    outraster2.save(outfile)

    del outraster, outraster2
    arcpy.Delete_management("FDI")
    return("mask clouds complete")