Bird's Eye ViewPosted in: Technology By Viviane Faria June 1 2014
Help, I can’t decide on the right drone!
LOOK NO FURTHER
One of the hottest new areas in precision agriculture is drones, or Unmanned Aerial Vehicles (UAVs). The technology is cutting-edge and exciting — and not just for its ‘boys and their toys’ appeal.
Progressive producers are already taking advantage of satellite imagery, Global Positioning Systems (GPS) and Geographic Information Systems (GIS) in their precision farming strategies. This has sparked the creation of new algorithms to monitor nutrient management, pest control, plant diseases, irrigation and yield prediction.
However, obtaining cloud-free images during the growing season has been a persistent issue. Temporal frequency of satellite imagery, the lack of sensors onboard satellites with optimal spatial resolution, and the processing time to obtain images has also limited research in this area.
Innovation in the UAV space is moving quickly to fill the gap, thanks to their ability to fly below cloud cover, when and as needed. According to a 2013 report by the Association for Unmanned Vehicle Systems International, precision agriculture accounts for 80 percent of the potential commercial market for unmanned aerial systems. This market is predicted to hit US$3 billion in the next three years, cresting to U$30 billion in the coming decade, said the report.
Growers are already using information coming from drones for drought management, disease detection, irrigation and pesticide application. UAVs provide producers with information about each area of the field, so they can water or apply inputs (nutrients, fungicides and herbicides) only where needed — saving money and reducing the farm’s environmental footprint.
UAVs are also being used by agrology consultants to gather data to improve the quality of advice and guidance they provide to their clients. In some cases, producers — primarily those running larger operations — are opting to buy the UAVs directly and incorporate the data into their overall management system.
Productivity, input costs and the environmental impact of farming are driving nearly all ag-related research and development today. As a result, the science is getting more complex and producers are being pushed to find new ways to perform field analysis in a timely manner to ensure that they can use that information to make the real-time decisions key to growing yields.
Drones are playing an important role — yet, as with any new gadget, even some of the early adopters are puzzled over which model to buy. Solid, unbiased reviews remain scant.
In an attempt to provide some objective, valuable analysis in this rapidly evolving space, we present the following comprehensive review of the top drones on the market today. The analysis was conducted by Viviane Faria, MSc in Remote Sensing, who works with Agri-Trend Inc.
Faria has worked in the Earth Observation Research Department of the National Institute of Space Research (INPE), a governmental organization in Brazil focused on basic research and technology development in the fields of space and atmospheric sciences. She focused in crop yield prediction, using agro-meteorological models, GIS and Remote Sensing, to provide the bases and guidance for the Governmental Policy in Agribusiness.
The following section is packed full of general information about UAVs, specifics on some of the top models and solid advice on how to decide which drone might be right for your farm operation.
Producers are using drones to provide three types of detailed views:
- Aerial imagery, which reveals patterns in the field caused by anything from irrigation problems to soil variation to nutrient deficiencies. It can even spot pest infestations that aren’t always obvious at field level.
- Multispectral images taken by onboard cameras capture data from the infrared as well as the visual spectrum, highlighting differences between healthy and distressed plants in a way that can’t be seen with the naked eye.
- Time lapsed images, in which a drone can survey a crop every week, every day or even every hour. These combined images create an animation that can reveal changes in the crop, highlighting trouble spots or opportunities for better crop management.
Researchers are also targeting drones. Studies underway on canola near Manitou, Manitoba are using NDVI (Normalized Difference Vegetation Index) to determine areas with high, medium and low biomass. Researchers then used this information to create zones for variable rate fungicide application.
In studying crop water stress, researchers found that images from either thermal or multispectral cameras were able to detect crop water stress in corn fields.
Using mean vegetation indexes (NDVI, GNDVI and LAI), nitrogen uptake and biomass maps have been produced for wheat fields in southwest France. Correlations over 82 percent were obtained when compared to ground-truth biophysical values.
The results of these studies indicate a tremendously bright future for integration of the spatially and temporally rich information provided through UAV imagery. The capabilities of management-oriented crop simulation models look increasingly strong.
In the end, to ensure they get reliable, readily usable information, producers should carefully consider the camera and platform they choose to buy. Pay attention to the sensor spectrum range, spatial resolution, noise correction during the flight time, and the methodology for your analysis.
Before You Buy
If you are interested in using a UAV, you need to understand what type of information you are looking for and then work backwards.
It’s important to understand what kind of information you want to obtain. This will determine whether you want to hire a service company or invest in your own drone. Isis Geomatics, NGF Geomatics, HighEye Aerial Imaging Inc., Precision Hawk, Farm Intelligence and Accuas are some options for flight and analysis services. Some of them have customised sensors, and some companies also offer specific analysis.
If you decide to purchase your own drone, it’s important to understand that the UAV is only a vehicle to assist in obtaining data. The camera is the most relevant part!
Simple cameras with three-four bands will give you an idea of what’s happening in the field, but the analysis will be limited. Consider cameras with spectrum filtering adjustment to obtain more reliable data. Multispectral (over 6 bands) such as Tetracam Mini MCA and hyperspectral (Rikola, Headwall), combined with LiDAR scanners and GIS, will give you much more insight.
The newly released cameras (S110 NIR and MultiSPEC 4C) could be another option for those who are looking for affordable options and better filtering capabilities.
At the end of the day, buyers’ need and price point will determine the ‘best’ camera. For example, the Tetracam Mini MCA is a good camera for the intermediate-level buyer, while those with hyperspectral sensors are better-suited for buyers looking for more advanced studies.
Picking the Right Platform
There are many options on the market today, all of which fall into two main categories: rotary wing and fixed wing.
Rotary wing units have the advantage of being able to hover and perform vertical takeoff and landings. Their battery doesn’t last long (since it’s used by the engine) and for this reason, they can only cover a small area. These UAVs are well suited to crop scouting or examining a particular area of a field.
Fixed wing units are able to fly at higher speeds for longer durations and can cover larger areas. Their disadvantage is not being able to hover and some require a runway or launcher for takeoff and landing.
For the most part, your choice of UAV will depend on the sensor, which will be determined by the specific information you want to collect. Once you pick a camera, find a UAV with suitable payload capacity (the weight the UAV can carry) to suit the sensor.
Most options have payload capacity up to 700 grams and have gimbals for specific cameras, which limits the sensor you can mount on them. These include eBee - SenseFly, X4ES - DraganFly, AgDrone - HoneyComb and PaceSetter - Precision Drone. Priced at US$10,000 to US$30,000, these models represent options for those who want to buy a drone without making a significant investment.
PrecisionHawk has a payload up to 1.3 kg and customizes their cameras (Hyperspectral and Lidar Scanners) to make them low in weight. QuestUAV has a payload of up to 2 kg, which can carry the Tetracam Multispectral MCA-6.
Some companies have developed platforms with even more payload capacity that offer flexibility in camera choice and the ability to integrate different sensors in the same flight. These include: Aibotix X6, Accuas Fixed Wing, ING Robotic Responder and AutoCopter G15. Priced at US$50,000 to US$72,000, these are all great options for a longer term investment.
No matter which platform you choose, keep in mind that you must obtain a mandatory Special Flight Operation Certificate (SFOC) from Transport Canada. And the UAV must to be operated within visual range.
Making Sense of Sensors
Choosing which sensor (camera) to buy is a key part of collecting good data. What the sensor captures is related to the interaction of the electromagnetic radiation (W/m2) and the target.
For agriculture, the spectrum range between 400 to 2,500 nanometers is very important. Green plant leaves typically display very low reflectance in visible regions of the spectrum (400 to 700 nm) due to the absorption of radiance by photosynthetic pigments.
Higher response occurs in 550 nm (Green band), which is the reason for the green color of vegetation. In the near-infrared regions (700 to 1,300 nm), the response is usually high due to the leave’s cell structure.
When foliage changes as the plant ages, or when plants undergo environmental stresses, leaf chlorophyll content declines. This reduces the green reflectance peak on the green channel, causing the tissues to appear yellowish. At the same time, Near Infrared (NIR) reflectance decreases, since the leaves structure has changed. Shortwave-infrared (SWIR) (1,300 to 2,500 nm) is related to the water content of the leaves as well as dry carbon compounds such as cellulose and lignin, nitrogen, sugars and other plant compounds that result when the leaf wilts and dries. Reflectance of leaves generally increases as leaf water content decreases. The SWIR band can be used in detecting plant drought stress.
The MaxMax Vegetation cameras, for example, have three channels (multispectral): Blue, Green and NIR (up to 770 nm). The Blue or Green channel is used to calculate vegetation index. Although they are the most affordable options, the filtering adjustment is not done properly, and there is spectral mixture meaning the Red Edge is included on NIR channel and NIR channel doesn’t include the peak of reflectance (around 850 nm). In order to improve this filtering, MaxMax has invested in a new camera (the S110 NIR) that covers NIR up to 1100 nm and keeps Red band, allowing the calculation of NDVI.
The MCA Mini Tetracam is also a multispectral camera that has a better filtering system. Each camera contains 4, 6 or 12 factory-aligned multispectral-sensors. Each sensor contains a customer-specified narrow-band filter that is inserted between the lens and sensor. One key advantage is the choice of up to 12 channels, which would help the producer generate very specific analysis.
Tetracam also provides Incident Light Sensor, which is another option to generate the reflectance data without the invariant calibration target. The cost ranges from U$12,000 to U$30,000.
The French MultiSPEC 4C is another option that was recently released by Airinov’s agronomy specialists and customized for SenseFly’s eBee Ag. It contains four separate 1.2-megapixel sensors providing image data over four highly precise bands (Green, Red, Red Edge and NIR). It also has an upward-facing irradiance sensor that automatically compensates for sunlight variations, resulting in increased data accuracy.
Hyperspectral cameras (they have narrow spectral bands over a continuous spectral range, which allows a detailed spectral signature of the target) such as Rikola, SphereOptics and Headwall are great options for deeper analysis. With hundreds of narrow bands along the spectrum, these cameras allow producers to study the vegetation spectral signature in more detail, and offer greater discrimination between crops.
Many vegetation indexes use specific bands and can only be explored with hyperspectral data. The high cost of these cameras — the Rikola is EUR €40,000, the SphereOptics VNIR 1600 is EUR €62,000 — remains a limitation.
Thermal cameras have also been explored for use in crop management. Water stress develops in crops when the evaporative demand exceeds the supply of water from the soil. When the plant is stressed and transpiration decreases, the crop canopy temperature tends to rise appreciably because of the reduction in evaporative cooling. For this reason, thermal infrared radiation is being used to monitor crop temperature. The presence of water and other gases in the atmosphere limits this measure to two wavelength windows, 3,000 to 5,000 nm and 7,000 to 15,000 nm. Thermal Eye 3600AS ($2,400 - $3,000), Nano Core-640 ($5,300 - $14,400) and Optrics PI400 ($9,800) are good lower-weight options.
To obtain detailed terrain information, a producer could use LiDAR scanners. These used to be very heavy, but lighter models (YellowScan, for example, is under 2kg and costs E70,000 including software) are now available.
Due to the high cost, many companies use stereo-photogrametry to produce Digital Surface Models (DSM), and the results have been satisfactory when the image is taken over bare earth.
Lidar can also be used for other things, including obtaining drainage, watercourse classification and forestry studies.
Finally, pre-processing is the last step before starting the analysis. You must correct the noise error that happens during the flight time and causes distortion in the image (pitch, roll and yaw axes influence, vignetting, lens distortion, etc). These can change the spectral, geometric and radiometric information of the image.
There is software available with these functionalities that typically runs for US$2,800-$6,000.
You will also need to choose a process for evaluating the data you have gathered — also known as your methodology. It refers to what kind of process will be used (pixel classification, spectrum signature algorithms, evaluations of the best vegetation index to be applied, regression analysis, etc) to analyze the information you have gathered.
Depending on flight time, sensor and methodology, you can quantify areas that weren’t germinated, discriminate between weeds and crops or identify areas that were affected by diseases or insects and see the impact. You can also cross-reference the results with field data and use the information for decision-making, which is the ultimate goal.
Now you’re ready to fly!
comments powered by Disqus