Growing & Winemaking

 

How I Learned to Stop Worrying and Love the Drone

December 2017
 
by Jim Meyers
 
 

I recently acquired a new hammer, in the form of an aerial drone, and have been using it to hover over vineyards to look for nails. I have found many “nails” in the process.

I, drone: Separation of concerns
In the past 20 years, the word drone has become synonymous with two types of devices: 1) multimillion-dollar remotely piloted war machines equipped with spy cameras and explosives, and 2) tiny four-propellered toy helicopters equipped with spy cameras piloted around your property by the neighbor's kids. Although many people think of a drone as either an unmanned airplane or helicopter, it is the lack of an on-board human pilot that defines the term, not the act of flying. A tank, for example, could also be a drone.

Perspective and freedom of motion
When collecting data via remote sensing, aerial vehicles have the advantage of altitude, which gives cameras or other sensors a very wide field of view. This enables them to inspect large acreages in very short timeframes. Compare this to ground-based vehicles, which must navigate obstructions and painstakingly zig-zag between crop rows to cover the same area. Small aerial platforms are hindered by short fuel supplies. If an aerial drone is used to perform a task at low altitude, it can no longer use a wide field of view to reduce task duration and will require refueling more frequently than a ground-based platform.

Who is really in control?
It is often assumed that drones are always piloted by a human via remote control, but this is not always true. While usually under the control of a human, drones are also always under the control of at least one computer and sometimes completely beyond the reach of human influence.

When planning a routine vineyard mapping mission, I use software on an iPad to define flight parameters such as the vineyard boundaries, flight altitude, preferred compass headings, camera settings and photograph frequency. When the mission is flown, the iPad is attached to the drone’s remote controller and the iPad does the actual flying. I only get involved when the software makes a mistake or evasive action is required.

That might be surprising to some readers, but the level of autonomy runs even deeper. When a human manually pilots a drone, the user employs a remote controller to actively direct the vehicle. But when a computer operates the remote controller, it has options unavailable to humans. Sure, the iPad could mimic human joystick movements to fly the drone from Point A to Point B, but it could also just tell the drone’s onboard flight computer to find its own way to Point A. This assumes that the drone is smart enough to do that, which is often the case. Once at Point A, the iPad can issue the next instruction telling the drone to find its own way to Point B.

The actual messaging between the iPad, remote controller and onboard flight computer is more complicated than that, but you get the picture. If everything is going smoothly, the drone is doing most of the thinking; the remote controller is sending messages to the drone that have little to do with joysticks; the iPad is intermittently barking orders at the remote controller, and the human pilot is standing around eating a ham sandwich. This layering of control is not so different from commercial air travel but at a vastly smaller scale in vehicle size, complexity, cost and special meal requests.

The highest level of drone autonomy occurs when the communication between the drone and the remote controller is lost. Choosing to abort the mission, the drone does its best to return to the initial launch location and land on its own while the defunct human pilot can only wait, hope and chew.

Freedom and consequence
Compared to ground vehicles, an autonomous aerial drone can claim the advantage of open space. When flying 400 feet above a vineyard, a 5-foot lateral deviation from the intended route has no consequence. That same deviation in a GPS-driven tractor will quickly wreck both vineyard and tractor. On the other hand, autonomous aerial vehicles require—both by law and common sense—constant human oversight. This prevents the human pilot from focusing on other tasks while the drone is working. Ground-based agricultural drones also have the advantage of being able to simply shut down in the event of an error without falling out of the sky.

What are drones good for?
To many, drones are considered to be most useful for photographing things and blowing them up—sometimes both, and not always in that order. This may be somewhat true for aerial drones, but they are also capable of carrying sensors other than cameras, and there are a few commercial examples of aerial drones equipped with small sprayers. In addition, a drone field tractor being guided by GPS can perform many tasks typically handled by human operators. Ground-based vehicles clearly have the overwhelming advantage for most day-to-day field operations.

Aerial drones in viticulture: an emerging case study
An eastern New York grower recently asked for a site inspection of a four-year-old vineyard block that was performing inconsistently and losing vines. Touring the block on an ATV provided a sense of broad inconsistency punctuated by areas of missing vines, soil erosion and low vigor. Each area of the block appeared to have different causes for the low vigor, but with some overlap. Organizing all that information from the ground to formulate an action plan can be tricky, so we flew the drone to make some maps.

The drone is equipped with two interchangeable cameras. One camera is an off-the-shelf red/green/blue (RGB) camera, which records photos and videos in the same format as a smartphone or point-and-shoot camera. The second camera has been modified to replace the blue light normally captured by a typical camera with near infrared (NIR) light. This is the camera used to create normalized difference vegetation index (NDVI) maps that measure plant vigor.

Visual inspection and vigor mapping with two-dimensional maps
The RGB camera was flown first and produced the map shown in the photo to the left. Viewed from above, large variability zones can be clearly identified in both the vineyard floor health and vine size. We plan to use this information to direct an intensive soil sampling investigation, starting by isolating soil samples in the large low-vigor feature in the lower portion of the block.

The flight was repeated with the NDVI camera and produced the three-zone vigor map shown on the image below. The zoned NDVI map helps identify areas of variability less visible with standard photography. Using this map, we can identify several additional smaller locations to sample in isolation.

At this point, you may be wondering what the USDA Natural Resources Conservation Service Soils Survey says about this block. Maybe it straddles multiple soil types? That is easy to check, because the aerial drone data automatically includes GPS information. A quick import into ArcGIS and an overlay of the soils survey data produced the image below. According to this map, there is only one soil type in the mix. If the block did straddle soil types, the spatial relationships between vigor and soil maps would be easy to reconcile.

Interactive 3-D inspection
One of the useful features of aerial mapping is the ability to create an interactive 3-D model of the vineyard. When mapping, the drone flies a programmed flight path over the vineyard and takes many overlapping pictures, which are later stitched into a large and detailed map. Overlapping images capture each point in the vineyard from multiple angles, which allows for a 3-D reconstruction. Anyone who has tried the View-Master stereoscopic toy has experienced the simplest form of 3-D photography.

Larger sets of multiple-perspective images enable models like that shown at right. The main image in this photo shows the spatial extent of the model. The lower inset shows a different view of the block captured by rotating the model in multiple axes and zooming in for detail. The upper inset illustrates an elevation map of the model.

To save flight time and processing time, this model was captured from an altitude of 350 feet with a minimal number of overlapping photos. That level of detail was high enough to interactively scout the block for the types of variables necessary for the inspection, but image and model detail can be substantially increased by flying at a lower altitude and increasing the image count. The photo below demonstrates a higher level of detail provided by a model captured using the same drone and camera at an altitude of 75 feet over a block of Pinot Noir in California’s Central Valley.

Getting started
Anyone considering the use of an agricultural aerial drone should be aware that the FAA recently declared strict rules governing their use in commerce. While children are still free to annoy their neighbors with very few restrictions, adults who want to use aerial drones for productive purposes are required to obtain an Unmanned Aircraft Systems (UAS) pilot certificate from the FAA and adhere to strict airspace rules and restrictions. Preparing for the exam requires a little determination, but there are some excellent study tools available that will ensure your readiness. A good place to start is the website faa.gov/uas/getting_started/. Feel free to contact me if you are interested in obtaining a license and/or have questions about the process.


Jim Meyers is based in Westport, N.Y., and serves as the viticulture and wine specialist for the Cornell Cooperative Extension in a 17-county region in eastern New York. He received his Ph.D. in viticulture from Cornell University and earned a master’s degree in computer science from Brown University.

 
SHARE »
Close
 
Currently no comments posted for this article.
 
 
SEE OTHER EDITIONS OF THIS COLUMN » CURRENT COLUMN ARTICLES ยป