Effektiv monitorering ved hjælp af åbne satellitdata
This presentation is based on some of the work done as part of my thesis entitled 'Monitoring Compliance with the Common Agricultural Policy' at Aalborg University Copenhagen.
Working with remote sensing data can be difficult, especially handling the vast amount of data generated by the 374+ operational earth observation satellites. This presentation shows a way of efficiently monitoring areas without the need of an enormous database or powerful computers.
After introducing the monitoring system, the presentation will move on to describe a use case of such a system to programmatically monitor compliance to agricultural regulation using open source machine learning tools.
Interesserede i monitorering af regulering, remote sensing-specialister, landbrugsentusiaster, Big Data-interesserede, udviklere og Open Source-interesserede.
Yderligere uddybning af abstract
On the 2007 March 7, the European Space Agency (ESA) launched their latest remote sensing satellite as part of their Copernicus Program. The satellite, called Sentinel 2B, joins its twin satellite, Sentinel 2A. The two satellites provide multispectral imaging services, and together they will supply imagery with a 10 m resolution and a revisit time of 2-3 days at latitudes between 30 and 60 degrees.
The launch of Sentinel 2B has spurred efforts to increase the uptake of the data provided in both the public and private sectors. In Denmark, we saw the Danish AgriFish Agency publish tenders to improve on their system to monitor the Common Agricultural Policy. EU National agencies pay out a vast amount of money in agricultural subsidies each year. A system to monitor effectively compliance at a large scale could be very beneficial to the EU as a whole.
The topicality of trying to assess the usability of the new data supplied by the Copernicus Program is the main inspiration behind this project. I do not strive to answer all the questions of the tenders nor to solve their problems directly; this work aims at creating a system for monitoring large areas using the new data supplied along with older yet still relevant satellites and creating a workflow to analyze the data using Machine Learning techniques, programmatically.
One of the most challenging aspects of working with remote sensing is handling the enormous amounts of data supplied. Thus, this research project is about handling this in a way that is meaningful and sustainable and creating both a user interface and an API for monitoring these areas. The second part of the project looks into creating a workflow for semi-automatic monitoring of some of the regulation specified in the CAP. The project delves into Machine Learning and remote sensing using a series of tools, such as the random forest algorithms contained in the Orfeo Toolbox, developed by the French Space Agency, CNES, to analyze the data from the monitoring system, programmatically.
The presentation will first present the monitoring system and give a quick ‘behind the scenes’ look into the inner workings. After that, I will move on to offering an example workflow and display some of the results from a test dataset from 2016. Following a discussion of the possibilities in the new datasets and modern remote sensing techniques.