This article features:

Jonathan Trotter, Technology Trials Manager Jonathan Trotter
Don Pendergrast, Technical Manager, Non-Combinable Crops

Advances in small-scale spectrometry devices, specifically those that can be integrated with drones, has enabled us to extend the capacity of our R&D programme through more detailed and accurate crop analysis.

Boosting Capability with AI

The intention was to use technology to perform accurate analysis of trial plots to compare the performance of different treatments, varieties and sites.

Having succeeded, we see that there is value to growers, especially as a tool for crop monitoring, harvest scheduling or for creating a bespoke and precisely targeted-cum-integrated crop management programme.

The system brings together off-the-shelf hardware with advanced processing software and a clever artificial intelligence program.

The program overlays several thousand images and the user then ‘trains’ the system by manually identifying several examples of whatever it is to be monitored. This could be crop of a pre-determined size, weeds or gaps between plants.

The intention was to develop a system that could be integrated with our R&D programme for more accurate and detailed crop analysis while overcoming the margin of error that comes from having different personnel perform the same role.

Once ‘trained’ the software analyses the whole image before producing a report based on the operator’s requirements. This could be to count whole cabbages or the number of heads that meet market specification of pre-determined size.

Similarly, it can be used to count and map weeds such as fat-hen or weed beet. This could then be exported to support a treatment plan for targeted herbicide application.

Proof of concept

The intention of our development trials in 2023 was to count cabbage plants to remove the need for these to be done manually.

In this example the trial plot was 0.16 ha and of the 5460 cabbages planted, 5241 were counted on 17^th October meaning a 4% loss.

Of greater value is the cabbage-by-cabbage analysis of head size to understand what proportion of the crop met market specification and if any portion of the plot needed to be left to grow on. The potential of such a system to inform harvest scheduling and yield forecasting is immense.

Ground-truthing the crop revealed a margin of error to be about 0.5 cm per cabbage. This is roughly about the same size of the picture resolution. Where greater accuracy is required, the drone can be flown closer to the crop though this has the downside of reducing working rates. Conversely, where less accuracy is required, the drone can be flown higher above the crop to increase working rates.

For trials, the data captured are extremely useful. We can compare different sites, treatments or whole regimes with considerable ease but also in immense detail. It can also be used to investigate possible reasons for underperformance, such as gapping between plants.

Having successfully proved the concept, the team is now using it in a sugar beet herbicide trial to count weeds and plants (for assessment purposes) while it is also being trialled on commercial farms as an agronomy tool.

The benefit to growers in applying this form of analysis to commercial crops is considerable and with more than 30 trained drone operators across the business we are well placed to deliver it.

Driving Gains in Productivity

Don Pendergrast, Technical Manager Non-Combinable Crops.

The technology represents a considerable opportunity for the industry to improve labour-use efficiency.

The adoption of technology has always driven gains in productivity, either through savings in labour or improvements in value. The use of drones offers both benefits. This technology was devised with our trials programme in mind, but it could just as easily be adopted by our customers too.