The Australian agriculture industry is powerhouse of food and fibrous material production in the Oceanic region, and is in a fantastic position to begin taking advantage of new technologies. Developments across a range of industries have paved the way for smaller sensors at a lower cost and weight than ever before.
It all starts with the humble smart phone. Consumer pressure for smaller, lighter and faster phones with better cameras and battery life have driven mass production of all of the involved components (Litwiller 2005). High precision GPS and accelerometers are on similar trajectories producing accuracies similar to larger systems. But why is this important?
Well, most if not all of these microprocessors, sensors and related systems are also involved in the development and production of miniaturised semi – fully autonomous unmanned aerial vehicles (UAV) and their payloads. After years of military development, commercial and hobby grade off the shelf systems or DIY kits are now ubiquitous, offering researchers, students, commercial operators and the interested general public the ability to build and fly fixed wing (think mini airplanes) or rotary based systems (helicopters but with more blades)
Payloads have also benefited, with new imaging and spectral sensors utilising camera technology and microprocessors developed for the smartphone industry. Importantly reductions in size have allowed bulky spectrometers to shrink to chipboard as big as your thumb (Hamamatsu Photonics, Japan).
So how will Australian farmers benefit? The above sensor is necessary for hyperspectral mapping, an imaging technique that records the entire spectrum of reflected light at each pixel. This spectral data can be analysed to produce important metrics for farming, such as soil moisture levels (Bajcsy & Groves, 2004), the ‘green-ness’ of the crop canopy which is related to the chlorophyll content of the leaves (Varshney, 2004) or modelling optimal harvest times based on crop yield (Gat et al, 2000).
To be usable by commercial farmers, the UAV platform, with it’s higher spatial resolution and greater temporal usability, needs to be a low cost alternative to traditional remote sensing methods for data acquisition. Space-borne and airborne sensors are already capable of producing this information at a much higher cost, and depend on cloud-free observations, and in the case of satellites, the ability to revisit a targeted area.
Lower altitude flying has great benefits for UAV systems, as the distance between target and sensor and resulting atmospheric effects are reduced while simultaneously allowing observations below the cloud layer. Spatial resolution needs to be of a similar scale to the targeted crops (cm to less than 1m scale (Uto, 2016)) to allow usable data on the individual areas of interest, which is only feasible for expensive airborne flights or the UAV platform. Further many environmental variations change through the growing season requiring frequent acquisitions, quickly accruing costs if airborne sensors are employed.
By producing these metrics for individual crops or areas, targeted ‘spoon-fed’ nutrient/pesticide strategies can be introduced (Bausch and Duke, 1996), rather than large scale spraying regimes. This benefits surrounding environments, by reducing harmful run-off from cropping areas into groundwater or river systems. Known as nutrient plumes, excessive nitrates and phosphorus levels promote algae growth in the system, in turn reducing the dissolved oxygen content in the system.
The process of identifying targeted areas was once a manual, labour intensive task of surveying individual crops in a field, however through the remote sensing of these metrics significant steps towards efficiency and automation can be made. Existing infrastructure (ie. paddock wide spraying systems), could be upgraded so that the remotely sensed data products produced by semi to fully autonomous UAV systems could be used to individual target affected areas.
This all sounds very complex, however, the commercialisation of UAV systems has also made great steps towards user-friendliness. Improved graphical interfacing, combined with microprocessors, known as flight controllers, has given users the ability to define GPS way points for UAV flight paths, automate take off and landing procedures, and fully autonomous data collection. This can take place in purpose built ground stations, from laptops, tablets or smartphone apps, depending on the complexity of the system.
Future agriculture UAV systems will improve known issues relating to hyperspectral data collection, namely the issue of coverage. Commercial hyperspectral mappers currently available, employ a type of sensor that is restricted to observations directly underneath the platform lowering the area covered per flight line. However, by utilising a similar method employed by terrestrial laser scanners, where a revolving mirror can direct laser beams around a stationary scanner, coverage can be improved (Uto, 2016).
Uptake of UAV systems into mainstream industrial agriculture will heavily depend on relative cost of platform compared to traditional methods and the benefits it will provided. Given the lowering cost of the platform and their payloads, their usability demonstrated by their temporal and spatial resolution benefits and ultimately the ability to characterise and automate environmental management decisions I believe it will only be a matter of time before mass uptake is realised.
Cover image: Flying agricultural drone. Creative Commons. available from: here
img 1: US Navy demonstration of UAV, Creative Commons. Available from: here
img 2: DJI Phantom 2, Lino Schmid & Moira Prati, Creative Commons. Available from: here
img 3: Mini spectrometer C11708MA, reproduced from brochure. Available from: here
img 4: Characterisation of potato field using hyperspectral imaging. Creative Commons. Available from: here
Bajcsy P. Groves P. (2004). Methodology For Hyperspectral Band Selection. Photogrammetric Engineering and Remote Sensing, 70(7), pp. 793-802
Hamamatsu Photonics. (2014). Mini-Spectrometers c11708ma. Available from: here
Gat N. et al. (2000) Estimating sugar beet yield using AVIRIS-derived indices. Summaries of the 9th JPL Airborne Earth Science Workshop. Jet Propulsion Laboratory, Pasadena, CA.
Litwiller D (2005) CMOS vs. CCD: Maturing Technologies, Maturing Markets. Photonics Spectra August 2005
Uto K. et al (2016) Development of a Low-Cost, Lightweight Hyperspectral Imaging System Based on a Polygon Mirror and Compact Spectrometers. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 9(2)
Varshney P. Arora M. (2004) “Advanced Image Processing Techniques for Remotely Sensed Hyperspectral Data”, Springer