Executive summary

Context

As climate change, food security, and sustainable land use become increasingly urgent issues in Europe, VITO and Superlinear recognized the need for innovative solutions to understand and monitor agriculture more effectively. They partnered to harness AI's potential to revolutionize agricultural monitoring, specifically crop classification.

Goal of the project

The objective was to develop a consistent, continental-scale crop classification system to support the European Commission's Common Agricultural Policy. This involved addressing the diversity of meteorological conditions and agricultural practices across Europe.

Solution

Superlinear created an AI model that integrated with VITO's existing tools, producing accurate, up-to-date crop maps for Europe. The model harnessed multispectral data for crop prediction using AI and deep learning technologies, offering a streamlined 'plug-and-play' experience.

Case study

The challenge

In today's rapidly changing world, understanding the intricacies of agriculture is more important than ever. As Europe faces the challenges of climate change, food security, and sustainable land use, new technologies and innovative solutions are required to keep up with the growing demands on agricultural resources. 

VITO, a research center with a specialized remote sensing unit dedicated to material, energy, healthcare, and sustainable land use, joined forces with Superlinear, an AI consultancy company. Their joint mission was to address the intricate task of crop classification and monitoring across Europe, leveraging VITO's expertise in remote sensing technologies to rise to this substantial challenge.

Together, they embarked on a journey to harness the power of artificial intelligence (AI) as a copilot in revolutionizing agriculture for the better.

Why is crop classification important? 

Crop classification is a vital process in the agriculture and food industry, enabling us to identify and categorize various types of crops on a larger scale. Accurate insights into crop type distributions are essential for supporting the European Commission's Common Agricultural Policy (CAP) and promoting sustainable practices.

Overcoming obstacles: embracing AI to address complex crop classification challenges

VITO faced a daunting challenge: accurately detecting and mapping crops throughout Europe, accounting for a vast area of interest and diverse crop growth patterns across different countries. The traditional methods of crop classification, which involved training separate models for each year and region based on available data, were insufficient to meet the demanding requirements of this ambitious project. 

The project required a more generalized approach, one that could adeptly handle the vast diversity in meteorological conditions and agricultural practices prevalent across different regions. The challenge was to develop a consistent, continental-scale product, despite the spatial and temporal bias in the training data, which meant ground truth was only known for certain times and locations in Europe.

To address this challenge, VITO sought to incorporate AI and deep learning (DL) technologies into their workflow, adopting the latest insights and developments in this domain. Recognizing the potential of AI as a copilot in their mission, they turned to Superlinear for a collaborative partnership.

The briefing

As part of a larger consortium, VITO won a public tender issued by the EEA, European Environment Agency, to create crop classification maps for use by the European Commission, local governments and private institutions. 

In the initial briefing, VITO and Superlinear discussed the challenges and discoveries that shaped their understanding of agricultural and remote sensing applications. They explored the potential of pixel-based crop classification for a sustainable future and the benefits of applying time series analysis to multispectral data. Together, they laid the foundation for a strong partnership focused on leveraging AI to unlock the potential of remote sensing in agriculture.

Ultimately, Superlinear was to create an AI model to be plugged into VITO’s existing tool to create accurate, up-to-date crop maps spanning Europe. The map will be one of the first continental products to be produced on the new Copernicus Data Space openEO API, which will become available to the broad public in July 2023. 

Thanks to more accurate crop maps, the end users will have a better view of crop distribution across Europe to make reports essential for food supplies, for example.

“The service of the European Environment Agency allows the deployment of VITO’s crop mapping expertise at a European scale. We are proud to deliver such a crucial map to the agriculture & food community”

Bart Deronde - Program Manager VITO

The work

A perfect blend of skill sets

The collaboration between VITO and Superlinear aimed to create a synergy that would push the boundaries of conventional crop classification strategies. Superlinear's AI and DL expertise complemented VITO's deep knowledge of agriculture and remote sensing, forming a team with the perfect blend of skill sets. Together, they adopted an iterative approach to crop classification, focusing on continuous learning and improvement.

The success of this partnership was built on effective communication and a shared commitment to innovation. Both teams could align their goals, discuss challenges, and make informed decisions by holding weekly in-person meetings and sprint sessions. This agile working environment allowed for rapid progress and a truly collaborative atmosphere.

“If we give Superlinear a challenge, we will have an answer in 1-2 weeks. We strongly appreciate that even if we book you guys for 1-2 days a week, you don't look at the clock and usually answer us within the hour. You never block us; you are very involved and always clear in your explanations.”

Kristof Van Tricht, Researcher VITO

Exploring the technical aspects

Throughout their collaboration, VITO and Superlinear delved into the technical details of pixel-based crop classification and multispectral data analysis. They addressed the challenges of data accessibility and inconsistencies in labeling systems across different countries. By exploring advanced Machine Learning techniques and optimizing the training process, they were able to create a powerful and efficient model for crop classification.

The AI-Driven Solution

The Superlinear team, leveraging their expertise in AI and DL, developed a model that utilized multispectral data to predict crop types. This model was a crucial component in VITO's workflow, allowing them to create accurate, up-to-date crop maps spanning Europe. The seamless integration of the AI-driven solution into VITO's existing infrastructure enabled a “plug-and-play” experience, streamlining the entire process.

“The collaboration between Superlinear and VITO exemplified the power of bridging the gap between domain and machine learning expertise. As we learned from each other's strengths, VITO's exceptional data preparation allowed us to focus on the seamless development of AI models, creating real impact fast and consistently driving performance further. This synergy resulted in a truly 'plug-and-play' experience, driving success for both organizations.”

Ruben Broekx, Team Lead and Solution Architect at Superlinear 

Interested in a full technical breakdown of how our AI team set up the solution? Read our blog article here. 

Impact

By embracing AI as a copilot, VITO and Superlinear have made significant strides in revolutionizing agricultural monitoring and reporting. Their solution provides a more reliable and accurate source of information on crop types and conditions, empowering governments and private institutions with the data they need to make informed decisions about agriculture and food security. 

As a result, the European Commission can implement the new Common Agricultural Policy, leveraging the transition to more efficient and sustainable agriculture.  

The future

The success of this partnership has led to a robust and ongoing collaboration between VITO and Superlinear. Their complementary expertise has opened up new avenues for innovation in the industry, with potential applications in the food sector, remote sensing, and beyond.

The VITO-Superlinear partnership demonstrates the power of AI as a copilot in solving complex challenges and driving innovation across various industries. As they continue to work together, their joint expertise in AI, DL, agriculture, and remote sensing will pave the way for a more sustainable, data-driven future for Europe and the world. By embracing the potential of AI and collaboration, VITO and Superlinear have set a new standard for agricultural monitoring and resource management.

"Combining VITO's domain expertise in remote sensing with Superlinear's top-notch AI skills may well be the ideal blend to tackle global challenges."

Kristof Van Tricht, Researcher VITO.