Hawking - Hummingbird: Space-enhanced machine learning to improve agrochemical management of crops

You are here

Growers around the world have unsophisticated tools to better manage their holdings and crops. Analysed satellite imagery is a scalable solution to help growers assess crop health, identify diseases, weeds and pest damage, create variable rate nutrient/chemical application maps, and help accurately forecast yields. This will allow growers to use targeted agro-chemical applications with the aim of reducing blanket agrochemical input, cutting farm costs and improving yields. Furthermore, it will mitigate the negatives of agrochemical overuse and disease resistance build-up.

Users and their needs

Benefit to users:
The following positive state of the art innovations are covered by the project, available to users of the free satellite service:

  • Disease Detection - Through our satellite offering, our platform supports the detection of specific diseases known to have significant impact on crop performance and yield, primarily:
    - Brown Rust
    - Fusarium
    - Downy Mildew
    Satellite data enables rapid collection of data over a more expansive set of spectral bands, providing additional dimensions to the datasets we are currently collecting. This leads to improvement of the accuracy of detection algorithms.
  • Weed and Fungal detection - In addition to disease, our platform identifies additional fungal disease infections and weed infestations such as Blackgrass and Downy Mildew before T2 derives from the satellite and historic data collected. Satellite data enables more efficient weed detection due to narrower band widths compared with commonly used multispectral sensors, this enables a higher spectral resolution and increased discrimination of the spectral signature related to weeds.
  • Crop Health Indicators - Satellite imagery used to monitor parameters such as crop height, weather forecasting and frost/drought (abiotic pressures). The scale of land covered by satellite data greatly improves the analysis of large-scale trends in monitoring the impact of weather on crop performance over multiple seasons, and on farms which have not previously been monitored through UAV data collection.
  • Field Management - Satellite data used to provide aerial observations on issues such as soil management zones, irrigation/drainage, crop grazing and headland/hedgerow distinction, all beneficial in the management of farmland.

United Kingdom, Russia, Ukraine, and Brazil.

Service/ system concept

The system is a cloud based web application which presents the farmer with meaningful, actionable data that it receives via the Hummingbird API. The system itself has been split into 4 distinct layers, or subsystems, that operate independently of each other with a clear data flow between layers. 

Space Added Value

The proposed project makes use of two Space Assets groups: Earth Observation and Satellite Navigation. By adding historical satellite imagery, in addition to real time data, to the platform Hummingbird is able to build an extensive time series of data onto which they can apply their proprietary machine learning techniques. The added dimension of satellite data as a deep, new layer of intelligence complements the platform by allowing them to:

  • Geo-reference UAV captured imagery during image acquisition
  • Avoid sensory issues and gather data in all weather (e.g. using SAR)
  • Increase the size of the machine learning training dataset for each specific, crop-related problem (thereby improving the overall diagnostic quality of the algorithms)
  • Perform atmospheric corrections (across satellite and UAV-mounted sensors)
  • Fuse UAV and satellite imagery into a high spatial and temporal resolution dataset

Current Status

The Project recently passed Site Acceptance Test milestone review and the pilot demonstration activities started April 5th, 2019 with UK farms. Satellite integration into Hummingbird platform has been completed. Next steps are to conduct the pilot activities and market the satellite offering to global agricultural customers.

Status Date

Updated: 12 April 2019 - Created: 20 June 2018