The growing world population and related food demand requires an intensification of our efforts to monitor and safeguard the worldwide agricultural landscape.
This starts by knowing what grows where! 

Our state-of-the-art crop mapping techniques are based on multisensory technology, where we combine complementary information from optical (such as Sentinel-2) and radar (such as Sentinel-1) sensors. 

A cropmap can be created in two distinct ways: pixel-based and parcel-based. Parcel-based methods have the advantage of much purer satellite signals, at the expense though of the challenge of having to delineate these parcels yourself.

With the aid of advanced deep learning techniques such as convolutional and recurrent neural networks, we are now able to fully automatically detect agricultural field boundaries and recognize the crop types on these fields based on joint Sentinel-1 and Sentinel-2 radar and optical imagery. More about this technology can be found in the article 'A Sentinel shake for early crop mapping'. 

This technology is useful not only for crop mapping efforts, but also for crop monitoring efforts, as explained in the article 'Looking through the clouds to improve crop monitoring'.

As a case study, we create a yearly cropmap for Belgium, although we constantly finetune and apply our crop mapping techniques for many different projects in regions around the world.