START Conference Manager    

Method for Automated Classification with INSPIRE Data and Sentinel-2 Satellite Imagery: Case Remote Crop Monitoring

Joona Laine, Sampo Savolainen and Jaana Mäkelä

(Submission #85)


Abstract

A method for remote crop monitoring with INSPIRE data and Sentinel-2 satellite imagery using machine learning methods is being developed for the use in Common Agricultural Policy (CAP) subsidy control. The model of the method could be trained and utilized with various kind of vector or raster data, e.g. with INSPIRE Land cover.

EU member countries are obligated to control the validity CAP subsidy applications. Each member country performs manual inspection for at least 5% of these subsidy applications. This is both expensive and a considerable administrative burden. According the European Union regulation No 809/2014, the crop type identification process in CAP should be carried out using remote sensing or orthophoto imagery for an alternative to physical inspections by competent authorities. Automated crop type identification would reduce the costs significantly.

With four bands of 10m and six bands of 20m spatial resolution and a maximum of 5 days temporal resolution, Sentinel-2 mission is a suitable source of satellite imagery for developing the identification method. The agricultural parcels could be extracted from the imagery using various segmentation algorithms and used in object-based classification or the spatial database with agricultural parcels could be used in parcel-based classification.

This paper will present an automated crop type identification method. Multiple different machine learning methods were trained and tested in the development process and the further development keeps up with the ongoing research of better machine learning methods. The agricultural parcels used in the training, testing and validation of this method were obtained from the Finnish Agency for Rural Affairs.

This identification method has been developed using the cloudy season of 2017 in Finland. Challenging conditions and the resulting low amount of cloudless imagery ensured that the developed method can be applied to many different use cases and climates. The further development and training of the machine learning model and INSPIRE and Sentinel missions make it possible to utilise this method in every EU country for the CAP subsidy control as well as numerous other tasks.

Categories

Topic Area:  [1.11] Agriculture - forestry - aquaculture
Abstract Type:  Oral Presentation

Additional Fields

 
Academic:   Yes
 
Data Provider:   Yes
 
Data User:   Yes
 
INSPIRE Implementer (IT):   No
 
INSPIRE newbies:   No
 
Policy Officers:   No
 
Public Administration (MS/Regional/Local):   Yes
 
Thematic specialists:   No
 
Comments:   Sentinel-2, Machine Learning


START Conference Manager (V2.61.0 - Rev. 5269)