Canopy cover data should handle missing data

[dlebauer]

Currently the canopy cover extractor runs on full field stitched images. However, it does not correctly handle missing data. Thus, on days with partial coverage the missing data weights the means toward 0, and we end up with plots in which the value of canopy cover is erratic, for example:

It doesn't appear that the solution will be as simple as deleting zeros after emergence, since some plots may have partial coverage.

The new image to plot aggregation will solve this issue, but in the mean time we need to remove data that is in error from the database.

@ZongyangLi is there a way to identify (and thus remove) data that is invalid without re-processing?

TODO

• [x] purge all canopy cover data generated by this extractor

  delete from traits where 
  variable_id in (select id from variables where name = 'canopy_cover') and 
  method_id in (select id from methods where name = 'Canopy Cover Estimation from RGB images');

• [ ] re-run extractor with changes that handle missing values

[dlebauer]

After discussing with @max-zilla today, it sounds like the canopy cover extractor can be re-run with existing updates that allow it to handle missing values.

[dlebauer]

Purged all previous canopy_cover data where checked > -1

delete from traits where  
    traits.variable_id in (select id from variables where name = 'canopy_cover') and 
    method_id in (select id from methods where name = 'Canopy Cover Estimation from Field Scanner RGB images') and 
        checked > -1;