Traceable acquisition and processing workflow for multispectral drone monitoring of orchards
This paper presents a traceable acquisition and processing workflow for DJI Mavic 3 Multispectral drone monitoring of orchards, detailing a comprehensive protocol from mission planning to quality control that was evaluated using data from durian orchards in Malaysia to enable reproducible data provenance and identify areas for future validation.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the heart of modern agriculture, a quiet revolution is taking place above the trees. Farmers and scientists have long relied on satellites to watch crops from space, but those views are often too distant to see the subtle differences between individual trees in an orchard. Ground-level scouting, where a person walks the rows, offers detail but lacks the big picture and is slow to cover large areas. Enter the drone, a small aircraft that can hover just high enough to capture sharp images of every tree, yet low enough to spot the specific health of a single branch. To make these images useful, they must be more than just pretty pictures; they need to be accurate measurements. This requires a strict set of rules for how the drone flies, how the light is measured, and how the photos are stitched together. Without these rules, a change in the color of a tree in a photo might just be a change in the weather or the time of day, rather than a real change in the tree's health.
A team of researchers in Sarawak, Malaysia, set out to build and test a clear, step-by-step guide for using drones to monitor durian orchards. They focused on a specific drone model equipped with cameras that see not just the colors humans can see, but also invisible light that plants reflect. Their goal was not to invent a new way to fly, but to create a reliable record-keeping system. They wanted to ensure that if someone flew the same drone over the same orchard next year, they could compare the results with confidence. The team looked back at three specific durian orchards they had already surveyed, checking their notes and data to see if they had captured enough information to prove exactly how the images were made. They found that while they could successfully reconstruct the entire journey from the flight plan to the final map, some crucial pieces of the puzzle were missing, preventing them from claiming the results were perfectly accurate measurements of the trees' true condition.
The researchers began by examining the flights they had already conducted over three blocks of durian trees, two in Sri Aman and one in Tapah. They used a drone that carries a standard camera for taking color photos and a special sensor that captures four different bands of light, including near-infrared, which is vital for understanding plant health. The team flew at different heights, ranging from 70 meters to 100 meters above the ground, depending on the size of the orchard block. At these heights, the camera could see details as small as 3.23 centimeters or 4.60 centimeters per pixel. They flew in a grid pattern, ensuring that each photo overlapped with the next by about 75 to 80 percent, a technique that allows software to stitch the images together into a single, seamless map. During these flights, the drone recorded the time of day, the speed of the wind, and the brightness of the sunlight, all of which can change how a plant looks in a photograph.
Once the data was collected, the team processed it to create several types of maps. They generated clear, color orthomosaics, which are like giant, high-resolution aerial photographs of the orchard, showing the layout of the trees, the dirt paths, and the surrounding landscape. They also created maps based on the invisible light bands, which allowed them to calculate vegetation indices. These are numerical values that indicate how green and healthy the plants are. For instance, they produced maps showing the Normalized Difference Vegetation Index, a common measure of plant vigor. The software successfully combined thousands of individual images into these large maps, and the team could trace every step of the process, from the raw photos to the final colored layers. This demonstrated that the workflow they proposed was functional and could produce the desired visual products.
However, the study also revealed significant gaps in their records that limit how much they can trust these maps for making serious farming decisions. While the team had the drone's internal records of the sunlight, they did not use a physical, calibrated white board on the ground to measure the light exactly. Without this ground truth, the numbers in the maps represent relative brightness rather than absolute, scientifically verified reflectance. This means that while the maps can show which trees look greener than others in a single photo, it is difficult to say for sure if a tree is truly healthy or if the difference is just due to a shift in the clouds or the angle of the sun. Furthermore, while the drone used a high-precision positioning system to know where it was, the team did not set up independent checkpoints on the ground to verify the exact location of every point on the map. They used the same points to guide the map and to check it, which is like checking your own work without a second pair of eyes.
The researchers concluded that their proposed workflow is excellent for organizing data and creating a clear history of how a survey was done, but it is not yet a complete scientific proof of the trees' condition. The maps they produced are valuable for spotting areas that need closer inspection, such as a section of the orchard that looks unusually brown or sparse. But the study explicitly states that these maps cannot yet be used to diagnose specific problems like nutrient deficiency, disease, or water stress with certainty. To do that, future surveys would need to include more rigorous checks, such as placing known light-reflecting panels on the ground and measuring the trees' health directly with soil and leaf samples. The study serves as a blueprint for what needs to be recorded to make drone monitoring reliable, highlighting that the technology is ready to fly, but the science of interpreting the data still needs a few more pieces of evidence to be fully trustworthy.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.