You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
Really should support area results better, being the reality that a lot of people multiply the count of pixels in a specific class with pixel size to get the area of X. For example, something like area of flooding in an area, by number of pixels.
Describe the solution you'd like
I'm not sure yet, but maybe options like
geoblaze.validArea(url,geom)// area of valid pixelsgeoblaze.invalidArea(url,geom)// area of no data and invalid pixels// like histogram, but instead of key => count, it's key => area (i.e. count * pixelarea)geoblaze.histogramArea(url,geom)
Describe alternatives you've considered
People could handwrite these calculations. We could also return metadata to the user to calculate the area themselves. Part of my concern is that if people use virtual resampling, that'll shrink the pixels sizes, so they'd have to use the dynamic pixel size of the virtual resampling calculation and not a fixed size.
I wonder if we could add frequency and area as params to the pre-existing histogram function
Is your feature request related to a problem? Please describe.
Really should support area results better, being the reality that a lot of people multiply the count of pixels in a specific class with pixel size to get the area of X. For example, something like area of flooding in an area, by number of pixels.
Describe the solution you'd like
I'm not sure yet, but maybe options like
Describe alternatives you've considered
People could handwrite these calculations. We could also return metadata to the user to calculate the area themselves. Part of my concern is that if people use virtual resampling, that'll shrink the pixels sizes, so they'd have to use the dynamic pixel size of the virtual resampling calculation and not a fixed size.
I wonder if we could add frequency and area as params to the pre-existing histogram function
Additional context
N/A
The text was updated successfully, but these errors were encountered: