CropCensus: AI Driven Crop Forecasting
Lead Participant:
FRUITCAST LIMITED
Abstract
FruitCast is developing cutting edge yield forecasting systems (yield and uncertainty; 6 weeks forward of any sampling) that fuse image analysis data of crop state (age, berry count, weight) with medium range ensemble weather predictions. This project, CropCensus, decreases time to market by at least a year and strengthens our first mover position in a highly competitive and innovative sector.
CropCensus's objectives are threefold; \[1\] crop centric data collection, \[2\] multi-site operational validation (data gathering, model accuracy, user co creation) of yield forecasts and \[3\] unlock technical barriers to scale;
Our approach is unique, FruitCast have already demonstrated unprecedented counting accuracy and scale (3.5M strawberries d-1). Competitor systems either \[1\] do not use image data, but are reliant on models that fuse meteorological data with human crop monitoring; \[2\] use static cameras of few berries that lack spatial data or \[3\] do not provide estimates of crop load (weight), critical for yield forecasts. No system has yet closed the loop between receiving image data and creation of accurate yield forecasts.
We will close the loop for strawberry yield forecasting, by:
\[1\] Extend image data collection system's functionality by onward developing our cloud platform to process data in real time allowing market wide deployment.
\[2\] Build our data bank to unlock scale via multi-crop/site (infrastructure tested on 3 sites, over 2 seasons)
\[3\] Site specific calibration of biophysical yield forecasting models integrated with weather forecast ensembles.
\[4\] Underpin scaling (and accuracy) by developing the digital architecture for forecasting at per-plant scale
\[5\] User co-creation of the reporting system, tested with live validation and feedback from growers.
The approach underpins a globally unique forecasting system, available as a service to growers. It meets the needs of an expanding F&V sector, where policy (National Food Strategy) aims to grow consumption by 30% in eight years. It de-risks the sector from critical labour constraints as planning is essential to manage depleted labour pools. Finally, it contributes to net-zero by reducing waste and supporting the digital transformation of food production.
CropCensus's objectives are threefold; \[1\] crop centric data collection, \[2\] multi-site operational validation (data gathering, model accuracy, user co creation) of yield forecasts and \[3\] unlock technical barriers to scale;
Our approach is unique, FruitCast have already demonstrated unprecedented counting accuracy and scale (3.5M strawberries d-1). Competitor systems either \[1\] do not use image data, but are reliant on models that fuse meteorological data with human crop monitoring; \[2\] use static cameras of few berries that lack spatial data or \[3\] do not provide estimates of crop load (weight), critical for yield forecasts. No system has yet closed the loop between receiving image data and creation of accurate yield forecasts.
We will close the loop for strawberry yield forecasting, by:
\[1\] Extend image data collection system's functionality by onward developing our cloud platform to process data in real time allowing market wide deployment.
\[2\] Build our data bank to unlock scale via multi-crop/site (infrastructure tested on 3 sites, over 2 seasons)
\[3\] Site specific calibration of biophysical yield forecasting models integrated with weather forecast ensembles.
\[4\] Underpin scaling (and accuracy) by developing the digital architecture for forecasting at per-plant scale
\[5\] User co-creation of the reporting system, tested with live validation and feedback from growers.
The approach underpins a globally unique forecasting system, available as a service to growers. It meets the needs of an expanding F&V sector, where policy (National Food Strategy) aims to grow consumption by 30% in eight years. It de-risks the sector from critical labour constraints as planning is essential to manage depleted labour pools. Finally, it contributes to net-zero by reducing waste and supporting the digital transformation of food production.
Lead Participant | Project Cost | Grant Offer |
|---|---|---|
| FRUITCAST LIMITED | £988,290 | £ 691,803 |
People |
ORCID iD |
| Raymond Kirk (Project Manager) |