Tuneable parameters for new segmentation methodology

[ellenbuckley]

I'd like to make sure these parameters can be changed or added to the config file:

• number of iterations
• erosion kernel
• area_lim variable- default is a function of the number of iterations

I'd like to make sure these options are available:

• save figs true or false
• which cloud mask to use- from modis cloud mask or fci nir band

[cpaniaguam]

• number of iterations

@ellenbuckley Does this correspond to 8 in the line below?

for n in np.arange(8,2,-1):

This produces a range array([8, 7, 6, 5, 4, 3]). Would it be good to have something like itmax (8) and itmin (3)?

[cpaniaguam]

• erosion kernel

@ellenbuckley
Is this a choice between kernel_er1=diamond(1) and kernel_er2=cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(5,5))?

[cpaniaguam]

• area_lim variable- default is a function of the number of iterations

@ellenbuckley
Currently the function is a power law. Are there any other functions to consider?

[cpaniaguam]

• which cloud mask to use- from modis cloud mask or fci nir band

@ellenbuckley

The code is "cloud mask source agnostic" -- it just grabs the files in the source directory, which is input/cloud in the notebook. Would the logic change if the files came from modis?

[ellenbuckley]

• number of iterations

@ellenbuckley Does this correspond to 8 in the line below?

for n in np.arange(8,2,-1):

This produces a range array([8, 7, 6, 5, 4, 3]). Would it be good to have something like itmax (8) and itmin (3)?

yes it does refer to the 8, it could be an array, because at one point i was playing around with the [8,6,4,2]

[ellenbuckley]

• which cloud mask to use- from modis cloud mask or fci nir band

@ellenbuckley

The code is "cloud mask source agnostic" -- it just grabs the files in the source directory, which is input/cloud in the notebook. Would the logic change if the files came from modis?

yes, the current cloud mask shows cloud mask as red where cloud >95% so that is why it pulls that value. In a new way we can use the reflectance data and just look a threshold in the first band of that to determine if there are clouds, so it would be a different process of interpreting that data

[ellenbuckley]

• area_lim variable- default is a function of the number of iterations

@ellenbuckley Currently the function is a power law. Are there any other functions to consider?

im not sure, ill have to think about this, and look at other code

[ellenbuckley]

• erosion kernel

@ellenbuckley Is this a choice between kernel_er1=diamond(1) and kernel_er2=cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(5,5))?

yes, can also think of others. I also at one point tried alternating between erosion masks