This is very interesting... I was just researching per-pixel classification, and remote sensing seems to be the field where that gets the most action.
I guess my question would be to any onlookers that would know better... is there any sort of python or Java or C# (or something scripty) library for general purpose per-pixel classification? Maybe on texture or colour?
This is pretty awesome, and I'm looking forward to playing around with it this arvo. Although I somewhat disagree with "Resolution at these bands is 20m per pixel, a lot less than the sub-meter imaginery of Google Maps, but enough for most crop fields."
From a whole farm overview point of view, 20m/pixel might be ok, i.e comparing entire paddock response, but for creating useful maps and applications in the field, we really do need better than that. We can individualize input control down to a per nozzle (plant) basis, so ideally the data we use to create maps should be as close as possible to that resolution.
Ah, I agree with you on that! This quarter Sentinel-2 should improve the freely available resolution to 10m per pixel, but sadly there aren't any open access higher resolution instruments (that I know of!).
However, note that "we can individualize input control down to a per nozzle" is sadly not prevalent everywhere! Even here in Spain the irrigation of a lot of crops dates back centuries and is basically flood irrigation, so the only control input is really "flood the entire thing, or not" :|
One thing which I find very interesting for high precision installations is the use of drones for automatically inspecting areas that need the most attention. I hope open hardware & open source drone platforms greatly improve the options for farmers!
6 comments
[ 3.0 ms ] story [ 21.3 ms ] threadI guess my question would be to any onlookers that would know better... is there any sort of python or Java or C# (or something scripty) library for general purpose per-pixel classification? Maybe on texture or colour?
From a whole farm overview point of view, 20m/pixel might be ok, i.e comparing entire paddock response, but for creating useful maps and applications in the field, we really do need better than that. We can individualize input control down to a per nozzle (plant) basis, so ideally the data we use to create maps should be as close as possible to that resolution.
However, note that "we can individualize input control down to a per nozzle" is sadly not prevalent everywhere! Even here in Spain the irrigation of a lot of crops dates back centuries and is basically flood irrigation, so the only control input is really "flood the entire thing, or not" :|
One thing which I find very interesting for high precision installations is the use of drones for automatically inspecting areas that need the most attention. I hope open hardware & open source drone platforms greatly improve the options for farmers!