Case studies by crop · Precision farming

Plant stand count: corn, sunflower, sugar beet

DJI Agriculture case study, collected by FLYDROPS.

Counting plants shows whether sowing worked and where action is needed, before it is too late. With a drone carrying an RGB camera and an AI platform, the study by DJI, Proofminder and the distributor Duplitec (Hungary) reports 90–99% accuracy, in hours rather than days.

For corn, the count is done when the plant has 3–7 leaves (V3–V7).

At a glance

Crops	Corn, sunflower, sugar beet (and other field crops, vegetables, trees)
Drone	DJI Phantom 4 Pro V2.0 or Phantom 4 RTK; RGB camera
Resolution	0.8 cm/pixel or better
Platform	Proofminder (AI)
Partners	Proofminder, Duplitec (DJI distributor in Hungary)

The challenge

Manual counting covers small areas and is laborious and inaccurate: the numbers do not represent the whole field and problem zones with uneven emergence or weeds go unnoticed. The result is wasted resources and decisions made on little data. An accurate count matters when checking sowing quality, estimating yield early or deciding on partial resowing.

Each plant is marked in the field with coordinates. Image: Proofminder.

What was done

The guide gives practical advice for capturing the images:

When: when the plant is visible from above but its leaves do not touch the next one. For corn, 3–7 leaves (V3–V7).
Height and speed: fly at 18–30 m, at 3–5 m/s.
Overlap: 75% front and side is a safe value.
Weather: stable, with little wind, and watch the exposure (overexposure is worse than underexposure).
Coverage: about 25–30 ha a day with a Phantom 4, if there are enough batteries; a 64 GB card for about 40–50 ha.

Common mistakes include wrong exposure, wind, too few batteries, flying after rain and inadequate planning for height differences.

The results

What the platform gives	Detail
Count accuracy	90–99% (per the guide), against a manual count that does not represent the whole field
Per-plant detection	Each plant is marked with coordinates; plant and row density
Additional findings	In the study’s corn, wild boar damage was found that the manual count had missed
Export	Shapefile export to compare with sowing data

The numbers come from the study and relate to that specific application.

In the same series, a second article shows the same method in sugar beet, at a large grower in Hungary, with 90%+ accuracy for spacing and gaps: read it on ag.dji.com.

What it means for Greece

Corn, sunflower and sugar beet are all grown in Greece, mainly in Thessaly, Macedonia and Thrace. The method is not spraying; it is mapping, and it relies on the same drone you can also use for multispectral analysis. Today our multispectral drone is the Mavic 3 Multispectral, and the Phantom 4 RTK for high positioning accuracy. The analysis in the study runs on an external platform (Proofminder); what matters here is how the images are captured.

Rules and responsibilities

The study is about fertilizing and mapping, not plant protection products. Drone flights are always subject to the competent aviation authority’s rules; check the requirements before you plan work in your own field.