Drone Based Pasture Managment
How It Works
Drone Imagery/Mapping
Our aim is to use drone imagery and AI learning algorithms to complement rising plate meter measurements.
The drone can map either the whole farm or selected paddocks to capture calibrated pasture imagery. We use a light reflectance calibration panel so that all imagery and data are referenced back to a fixed standard. This calibration step is the first and most important part of the process, as it helps ensure consistent and accurate pasture cover predictions.
Drone imagery must be captured under consistent ambient lighting conditions. Suitable conditions include fully overcast skies, clear skies, or stable scattered cloud, provided the lighting remains consistent across the entire mapping mission. Each image in the mission needs to be exposed to similar light conditions to maintain accuracy.
Images affected by passing cloud, sudden sun breaks, or inconsistent lighting can be filtered out before processing. This allows more practical opportunities to map and measure pasture cover while still maintaining reliable data quality.
In perfect conditions, a drone can map 200 hectares in 25 minutes.
Rising Plate Meter
Invented in the 1970s, the rising plate meter remains one of the most reliable tools for estimating pasture dry matter in a paddock.
Our drone imagery and AI algorithms are designed to complement, not replace, rising plate meter readings. All drone-based predictions are backed by ground-truth measurements collected using a rising plate meter.
The system provides the raw compressed pasture height value for each paddock. From there, farmers can apply their own dry matter equation to calculate the final pasture cover result that best suits their farm, pasture type, and management system.
Ground Truths
We compare rising plate meter ground-truth results directly against drone imagery captured over the exact same area.
Our ground-truth areas are 10 m x 10 m squares, marked using GPS. Each square is comprehensively measured with a rising plate meter to calculate an accurate average compressed pasture height.
This ground-truth result is then matched against the calibrated drone imagery from that exact 10 m x 10 m area. The data is stored in our calibrated ground-truth library, allowing it to be used later for AI training, comparison, and future pasture cover predictions.
Each ground truth is also labelled with the pasture condition at the time of measurement, such as ryegrass/clover with seed head, topped pasture, grazed pasture, baleage aftermath, young regrowth, or other relevant stages. This allows the system to learn how pasture appears under different conditions, growth stages, and management situations.
Training Our Ai
The system first builds a master model, where every ground-truth and drone-image data point is stored. This provides a broad reference library across different farms, pasture types, seasons, and growing conditions.
The model also considers factors such as pasture state, time of year, pasture species, and management history. As more ground truths are collected on a specific farm, the system can begin building a farm-specific model. Over time, this can be refined further into paddock-specific models, and eventually into models that account for different zones within a paddock and their unique characteristics.
The system can also identify and remove non-pasture areas from the prediction, such as bare dirt, weeds, tracks, shadows, and other features that may affect accuracy. This helps provide a more reliable pasture cover result while allowing the model to adapt to the individual characteristics of each paddock.
Predictions
Predictions are made by breaking each paddock down into 1 m x 1 m grid squares.
The system works through each square individually and compares the calibrated drone imagery against the ground-truth library, rather than simply generalising the paddock as one whole area. This allows the AI to generate a more detailed and accurate compressed pasture height prediction across the paddock.
As the model learns, it adapts its approach based on the farm, paddock, pasture state, time of year, and previously collected ground truths. Where enough data is available, the prediction can be guided by a custom farm-specific or paddock-specific AI model.
If the predicted pasture results do not match the real paddock conditions closely enough, additional rising plate meter ground truths can be collected and added to the library. This helps the system better understand the paddock’s unique characteristics and improves future prediction accuracy.
Outputs
Our prediction approach allows for a wide range of practical outputs.
Because each paddock is broken down into 1 m x 1 m cells, the system can generate highly detailed pasture maps rather than a single paddock average. This means we can request an exact grazing break layout designed to create even feed allocation across each break.
By inputting the herd’s daily feed requirements, the system can help build a grazing plan for that paddock. This has strong potential for use with modern virtual fencing systems, where accurate feed allocation and break placement are critical.
Actual grazeable area and available feed can also be determined by comparing pre-grazing and post-grazing predictions. This helps identify how much pasture was realistically available and how much was actually utilised.
Over time, this information can influence future grazing decisions, improve pasture management, and provide valuable data for variable-rate nutrient application.
Why Use A Drone?
Drone and AI pasture measurement is designed to complement proper hands-on pasture management, not replace it.
Good pasture management still relies on farmer knowledge, stock sense, paddock history, and accurate ground-truth measurements. Our system builds on that by using rising plate meter readings, farm observations, and calibrated drone imagery to continuously improve the model over time. The more ground truths collected, the better the system becomes at understanding each farm, each paddock, and the different pasture conditions within them.
The real advantage is speed, scale, and detail. Instead of relying only on a handful of manual measurements across a farm, the drone can rapidly capture high-resolution imagery across whole paddocks or the entire farm. The AI then converts that imagery into detailed pasture cover information, helping farmers make better day-to-day decisions around grazing, feed allocation, growth rates, and pasture utilisation.
Unlike satellite-based systems, this approach is not trying to build predictions from low-resolution or inconsistent data. The drone captures eyes-on, properly calibrated imagery at extremely high resolution, under controlled and selected lighting conditions. This gives the system a much clearer view of what is actually happening on the ground.
In practical terms, it is like having the ability to plate meter every square metre of the farm regularly, without the time and labour normally required. This allows farmers to keep a much closer eye on pasture growth, identify feed shortages earlier, track paddock performance, compare pre-grazing and post-grazing covers, and stay ahead of feed management decisions.
Because the system is trained from the farm’s own data and observations, it becomes more relevant and reliable over time. It learns how that farm’s pasture looks under different conditions, seasons, grazing stages, and paddock histories. This creates a pasture measurement tool that supports the farmer’s experience while providing fast, detailed, and repeatable data that can be used for grazing plans, virtual fencing, feed budgeting, fertiliser decisions, and long-term pasture improvement.