Right description.
Unsupervised detection of plantations remotely in satellite images.
Actually a lot of finance contracts around agriculture needs some proof of existence. To go to the place
of that plantation are declared to exist would be slow and expensive. What if we can detect plantations using
georeference + machine learning? This is what I’m going to show using Iris as a database and computing server.
For this experiment I use images from the Satellite Sentinel (https://sentinelhub-py.readthedocs.io/en/latest/)
which provide a time series of georeferenced images.
As the raw data has 121km² of area and I dont need to process all this area, I crop the desired area into a smaller
area square (around 1km²).
After all the image processing we can calculate the vegetation index NDVI (Normalized difference vegetation index) to
understand if the area has or not something growing.
Putting this solution into a continuous data pipeline can help to remote monitoring of growing vegetation areas.
In fact the new Iris feature to use Python Embedded make this kind of solution to be quickly created into
the IRIS environment.
First clone my repo in a directory of your machine:
# git clone https://github.com/renatobanzai/iris_satellite_plantation.git
After this, go to the directory iris_satellite_plantation
# cd iris_satellite_plantation/app
# docker build -t iris_satellite_plantation:latest ./
Sometimes the command above takes longer than the reality. Just do a ctrl + c and repeat this step.
# docker run --name iris_plantation -d --publish 1972:1972 --publish 52773:52773 iris_satellite_plantation:latest
# docker exec -it iris_plantation /usr/irissys/bin/irispython /app/process.py
Now you are officially running the newest EAP, connecting into AWS to download a satellite image and much more soon ;)
https://openexchange.intersystems.com/package/iris_satellite_plantation