Detecting the world's most accurate field boundaries

Informacje na temat projektu

CropCloud

Identyfikator umowy o grant: 190112197

DOI

10.3030/190112197

Projekt został zamknięty

Data podpisania przez KE 6 Października 2022

Data rozpoczęcia 1 Czerwca 2022

Data zakończenia 31 Maja 2024

Finansowanie w ramach

• The European Innovation Council (EIC)

Koszt całkowity

€ 2 531 250,00

Wkład UE

€ 1 767 500,00

Koordynowany przez

DIGIFARM AS
Norway

Periodic Reporting for period 2 - CropCloud (Detecting the world's most accurate field boundaries)

Okres sprawozdawczy: 2023-06-01 do 2024-05-31

Podsumowanie kontekstu i ogólnych celów projektu

All precision agriculture services and in-field analytics starts with accurate field boundaries and seeded acres. Unfortunately, we’re currently making critical decisions based on inaccurate and outdated data which is affecting the entire agricultural value chain.

Existing solutions for large scale field boundary data consists of Cadastral map data (national agricultural paying agencies, LPIS in 27-EU regions and CLUs in the US) which are outdated, inaccurate and time-consuming to update and maintain as this is done through manually drawing field boundaries into a digital map-solutions. There are ~34.6 million field boundaries in the EU-regions and 32 million in the US (not updated since 2008); a 5 hectare field takes one person approx. 1 minute to manually draw; 66.6 million boundaries would take 125 years.

Norwegian start-up DigiFarm has developed a deep-resolution AI-model for Sentinel 2 imagery to increase image resolution from 10m to 1m, combined with a deep neural network model that automatically detects field boundaries and seeded acres to power precision agriculture. The EIC-funded CropCloud project will exploit the latest advancements in AI technology and super-resolution of satellite data to improve detection algorithms to achieve 12-15 % higher accuracy than existing solutions. This reduces both the cost and time of current manual digitisation by ~99%.

Prace wykonane od początku projektu do końca okresu sprawozdawczego oraz najważniejsze dotychczasowe rezultaty

We have successfully completed the following deliverables according to the project WPs:

D1.1. Updated Financial Forecast and Strategy Report
D2.1. Prototype demonstration of web- and mobile application (MVP)
D3.1. Performance assessment report from initial trials

Additionally, we have also completed the following key achievements:

1. Developed technology for automatic delineation model accuracy of 0.94 IoU metrics, i.e. 94% across all pilot regions: United States, Canada, Brazil, Myanmar, Thailand, India, Vietnam, Germany, Italy, Spain and France.
2. Successfully developed API endpoints to match pilot users and potential clients demand and requirements, i.e. including (a) single point (b) bbox and (c) coverage API.
3. Successfully developed front-end web-application, beta tested and launched MVP to pilot users.
4. Successfully validated and commenced commercial expansion: since the project started (June 2022) we have delivered 3.5 million boundaries to over 7 corporate clients and secured 235k EUR in revenue-generating contracts and reached over 350k farmers (end-users) and built up a sales pipeline of 325 leads and €6.7m funnel.
5. Secured POCs from 42 clients in 17 countries including B2B/B2Gs such as: Bayer Crop Science, KWS, Limagrain, Yara, FMC, Agmatix (ICL Group), Monsanto (FarmRise), Mahindra & Mahindra, CNH Industrial, National Paying Agency of Lithuania (Govt), National Paying Agency of Austria (Govt), USDA etc.

Innowacyjność oraz oczekiwany potencjalny wpływ (w tym dotychczasowe znaczenie społeczno-gospodarcze i szersze implikacje społeczne projektu)

DigiFarm's solution provides benefits to both end-users: farmers, growers and agronomist as well as directly to B2B/B2G entities who conduct agricultural analysis or provide digital agricultural services, this includes:

1. Reduction in crop input costs: crop input accounts for 50% of total crop production and it’s a widely known fact that 40% of the world's agricultural fields are over fertilized and farmers are on average losing 10-15% in yield potential through inadequate input application. DigiFarm provides services to our client network which enable growers around the world to reduce their input costs by up to 15% and increase yield potential by up to 10% through variable rate technology (VRT) input optimisation (fertilizer and fungicide). DigiFarm has delivered this service to KWS, one of the largest seed-producers in the world, where DigiFarm helped KWS onboard and digitize 70,000+ farmers for KWS (MyKWS) enabling them to provide actionable insights to farmers across 4 countries, reducing crop input costs up to ~15%, by providing in-app variable rate seeding maps.

2. Reduce cost of digital farming services and SatEO imagery at the same time as increasing adoption of precision ag-services: we have successfully developed a solution which will enable to address all these 3 points which have been significant barriers and challenges in the digital ag-market including:
a. Reduce cost of creation and updating field boundaries by 99% through providing automatic solution vs manual delineation
b. Increase accuracy of boundary delineation (+ seeded acres) by 12-15% through providing an automatic solution which achieves higher accuracy than manual delineation, due to several factors including: up-to-date high-res SatEO basemap and extensive training dataset.
c. Reduce cost of high-resolution SatEO imagery by 75% through providing super-resolved Sentinel-2 at 1m per pixel to clients, reducing the barrier in the market, compared to existing solutions such as Planet, Airbus, Maxar etc.

Below is a table which highlights the difference between current issues and our solution below:

Data Layer Existing Approach DigiFarm’s Solution
Satellite Data EO If we were to buy commercial grade SatEO data it would cost us €1m/year to buy input data for Germany alone. It costs us €10k in (GPU/CPU) processing for all of Germany and we can do it instantly in 24 hours.
Field Boundaries It would cost approx. €6.3m/year to manually draw and digitize all the boundaries in the US for example (175m ha’s) and would take 15 people, working full-time for 1 year to do it. Currently, we can automatically delineated field boundaries at 12-15% higher accuracy across all of the US in 4 days with processing costs of €20k

Figure 2. illustrating automatically delineated field boundaries using SatEO data.

Figure 4. illustrating the full API documentation enabling various endpoints for integration.

Figure 3. illustrating field boundaries delineated in one of the pilot client areas of interest.

Figure 5. illustrating the front-end interface (web-app) of the API Sandbox serving the various endp

DigIFarm’s illustration for it’s technology, i.e. delineating field boundaries and it’s stacked data