Training to be a horticulturalist, you probably did not expect that one day you may consider gaining an “aviation” certification to care for plants. Some production nurseries are doing just that, using drone-mounted cameras to capture imagery to manage plant count and health.
by Gabrielle Stannus
Drone-mounted cameras can provide high spatial resolution and fast turnaround capabilities whilst remaining relatively low-cost and easy to use when compared to satellite imagery.
Manufacturers claim that data collected from drone-mounted cameras can be used to understand nutrient levels and determine fertilizer application; plan drainage and irrigation; compile plant counts and analyse stand establishments; reduce crop damage; and detect fungi and pests.
Drone-mounted cameras are especially useful when it comes to plant inventory. Plant inventory management still mostly involves labour and time-intensive manual counting, especially for orchards and nursery fields with large areas, with those growers often counting only a portion of their crop, and inaccurately.
This is an issue shared by many tree growers, including J. Frank Schmidt and Son, a wholesale production nursery based in Oregon, USA. This business grows 500 shade and flowering tree varieties over 2500 acres, employing more than 350 full-time staff and 150 seasonal/casual workers.
Sam Doane, Production Horticulturalist at J. Frank Schmidt and Son, says that with a few million trees sold every year, even a small error percentage is more than what you would like.
Sam looked to drones almost a decade ago to help him develop a more accurate inventory of tree stock. To take his first pictures of the nursery fields from the air, Sam used a ‘point and shoot’ camera attached to a helium balloon. “All I got was a bunch of out of focus pictures”, Sam laughs.
Drone, camera, action!
J. Frank Schmidt and Son now fly two drones: the DJI Phantom Pro 4 and the DJI Matrice 100. The nursery uses the DJI Phantom Pro 4 mostly for marketing purposes. Sam flies the DJI Matrice 100 mounted with one of either two cameras: one in Red-Green-Blue (RGB) and the other Near Infrared-Green-Blue (NIRGB).
Sam uses the DroneDeploy app to help plan, launch, conduct and land his drone missions. While the drone lands itself at the end of the mission, by law, Sam must remain in visual line of sight (VLOS) of the drone at all times.
Sam then uses an Agremo app, linked to his DroneDeploy account, to upload and store images collected from the drone’s camera to Agremo’s website. There Sam can stitch images together, analyse them and generate reports. Sam finds this app useful for counting discrete plants, but not so good when canopies overlap. And the current charge, $1.50 (USD) an acre, with a ten acre minimum, is not cheap.
“One acre of a variety is a large planting for us”, Sam says, adding that he can still manually count more cheaply. Sam suggests to other growers interested in putting up drones that they “add value in the sky, start with something affordable”. For J. Frank Schmidt and Son, that means getting a return on investment with marketing images.
And now for some maths!
Once the maps are uploaded, some mathematics is required to make them meaningful. Cameras capturing Near Infrared (NIR) light can apply the Normalized Difference Vegetation Index (NDVI) and other algorithms. These algorithms compare the proportions of light captured across the different bands they register and then compute numerical values for each pixel or area of a given drone map. These maps are then assigned colors based on those numerical values, making it easier to identify between healthy and unhealthy areas.
Normalized Difference Vegetation Index (NDVI)
NDVI nominally indicates the health of green vegetation. Plants appear green to humans because chlorophyll strongly reflects NIR while red and blue are absorbed. A healthy leaf’s spongy mesophyll usually reflects most NIR. In dehydrated or stressed plants, this spongy layer collapses, meaning the leaves reflect less NIR light than usual, but the same amount in the visible range. Thus, mathematically combining these two signals can help differentiate plant from non-plant and a healthy plant from a sickly one.
Sam suggests that this technology is currently more suited to monocultural ‘green’ crops such as corn and soybeans. He urges caution when interpreting the results of these algorithms. “Believe it or not, green is not always the colour we want” says Sam. Maples come in many shades of green and there is even a Yellow Leaf Maple. An uneducated user of this technology may infer from the NDVI results that this species requires fertiliser or is drought-stressed, when it is just its natural healthy colour.
NB. If flying a drone in Australia, you must comply with the Civil Aviation Safety Authority’s (CASA) drone-flying rules and regulations.
Potential application in retail garden centres
Amazon and convenience store chain 7-Eleven have been trialling drone technology since about 2015. However, Dr Louise Grimmer from the University of Tasmania’s Tasmanian School of Business and Economics believes we are unlikely to see drones in retailing in any meaningful way over the next decade or so, particularly in Australia and New Zealand. Issues around privacy, security, logistics and safety, as well as the limits on weights that drones can accommodate, need to be worked out first.
Drones are unlikely to be used in delivery services any time soon, given the cost and need for operators to remain in the visual line of sight (VLOS) of the drone in operation.
“What we might see in the shorter term is the larger retail garden centres using drones in their stores and onsite, rather than flying from stores to customers’ doors. Drones with video applications might be used to assist with security checks, securing warehouses and checking stock numbers for re-ordering and customer queries, and for helping with deliveries to customers’ vehicles” Dr Grimmer says.