The basic idea of precision agriculture is the knowledge of the land heterogeneity, i.e. information on the composition and quality of soil and vegetation in individual locations and soil blocks. Unmanned aerial vehicles with special sensors are an ideal solution for high-precision mapping of soil and crop conditions on a regular basis and in a short time. In contrast to satellite photographs, they provide up-to-date data without having to wait for the passage of a satellite and a suitable cloud situation. As compared to aerial imaging from a manned aircraft, the advantage of UAV is fast readiness for operation, even in bad weather conditions, and significantly lower flight costs. Moreover, drones can fly at low heights and, therefore, collect data at a resolution of up to 1cm/pixel, which, so far, cannot be achieved by any other method.
Unmanned aircrafts of our own design have a long flight time and, therefore, they can also be used to cover large plots of land. We use multi-rotor drones if a high resolution of data are required. We use the well-proven Parrot Sequoia sensor for multispectral imaging which was designed to be integrated in unmanned aerial vehicles used in agriculture. Optionally, we also carry out visible-spectrum imaging for precision land mapping. Drones can also carry a thermal camera in order to detect animals before mowing and analyse irrigation and water stress of plants.
All multispectral images are saved along with spatial coordinates. Then, special software is used to make a map of the photographed field with visualization of the required indicators and selected vegetation indexes. Another output is application maps for ground-based variable-dosing machinery. We analyse the data in mutual cooperation with the Crop Research Institute in Prague and receive from them data evaluated to high precision and well-researched conclusions with regard to variable fertilizing.
Another advantage of our solution is the processing of collected data and a possibility for our clients to view and inspect them using our SkyView online user interface. After logging in, clients can view the inspection results from anywhere using their devices, without having to download the entire data package. No specialized software is needed to perform the basic analysis. The browser interactively shows individual spectral components and index maps and allows measuring of lengths and areas, including GPS localization.
Below, you can interactively view the data collected from one of the experimenting fields during a test conducted by the Prague VÚRV institute. The institute's employees prepared a testing environment on 96 parcels (9x9 m) of sugar beet. There were 24 alternatives: 6 mineral-fertilizing levels (nitrogen 0-130 kg per ha) and 3 kinds of organic fertilizing (manure, slurry and compost), including alternatives with no organic fertilizing. The processed data were related to the amount of actually harvested sugar beet and the analysis showed a close relationship between the identified spectral characteristics of plants and the yield of both the aboveground and underground biomass. The experiment proved that photographs taken by a drone equipped with a Sequoia multispectral camera can be used to precisely predict the yields of sugar beet bulbs relative to the nutritional status of soil.
The NDVI index is shown on the right hand side. Vegetation indexes are generally used for vegetation mapping and for plant quantitative indicator monitoring based on spectral reflectiveness. On the left hand side, you can see standard photographs in a visible spectrum and at a resolution of approx. 1.2cm per pixel, for the sake of comparison. The graph shows a correlation between the NDVI index and the actual amount of harvested biomass.
The visualization can be moved using a cursor and the scale can be changed using buttons.
The basic principle is the knowledge of spatial and temporal variability of properties of soil and vegetation on an agrarian land and application of such data to render farming more efficient. In the past, farmers knew their fields. They knew what parts gave high yields and, on the other hand, where the productivity was lower and, thus, the work done did not pay that much. However, they could not effectively get such knowledge after the areas of land and sizes of agricultural farms had grown much larger globally. The land was rather treated as a homogeneous area and its potential could not be fully used. The situation only changed with the availability of technology and necessary technical equipment. Today, they make it possible for us to obtain and analyse a wide range of necessary data from multiple sources. Their outputs help farmers decide on farming activities, adjust variable application of fertilizers and pesticides in proper amounts and in the right places and, last but not least, predict the conditions and properties of soil and vegetation. It is Unmanned Aerial Vehicles which bring the possibilities of precision agriculture another step forward.
Unmanned aerial vehicles started to be used in agriculture and their advantages were already known in the past when the first drones became available to the general public. However, this technology has only recently reached a point at which it can be considered a reliable device of everyday use and cooperate with other technical equipment of the farm.
Drones are primarily used for multispectral imaging and mapping of land. Currently, such data are also received from other sources, such as satellite images or images taken from manned aircrafts. Satellite data cover a large area at the expense of their lower accuracy. However, the greatest disadvantage has so far been the fact that the images are old and out of date as a result of the interval of satellite passage over a particular area (moreover, depending on the cloud situation). Manned aircrafts provide higher-resolution data which can be collected in a relatively short time after placing a request. However, such method is associated with high flight costs. On the other hand, drones have a little disadvantage as compared to the above typical methods of data collection: their shorter flight time. When monitoring a large area of land, drones have to land so that their drive accumulators could be replaced. But battery technology keeps developing and modern vehicles of suitable design can already remain aloft for several hours.
The greatest advantage of drones is their effectiveness in terms of agricultural inspection: the preparation only takes minutes and the most suitable flight parameters and sensors can be precisely defined and chosen for each flight. After entering the area of destination and imaging parameters, an optimum flight plan is chosen and the entire flight is performed automatically. The flight plan can be saved and re-used at any time in order to repeat imaging in certain time intervals. The flight is only limited by the weather conditions, but modern drones can fly in visibility and weather conditions which would present an unnecessary risk for a manned aircraft. Thanks to such properties, including the option of immediate image preview, UAV provide the most up-to-date data of the given land and make it possible for the farmer to decide on necessary actions in a timely manner (which is even more important in the event of field infestation by pests or natural disasters). Moreover, UAVs can fly low and take photographs at a resolution of up to 1cm per pixel which makes it possible to quantify individual plants, accurately delineate vegetation and plots of land, and locate weed and parasites.
An unmanned vehicle is ready for flight within a few minutes. Up-to-date data are available within a few hours thanks to autonomous flying and automatic processing of photographs.
In contrast to satellite images and mapping from a manned aircraft, a drone can collect data at a resolution of up to 1cm per pixel thanks to its low flight.
High-resolution images provide new processing possibilities which have been unthinkable until recently. For example, we can use image processing algorithms to obtain data on the number and height of individual plants.
Typical spectral characteristics of plants are used during imaging, i.e. different levels of reflection and absorption of individual components of the light spectrum. The reflectiveness of the blue and red components is very low because this band is absorbed due to photosynthesis. The green component has a higher reflectiveness and that is why chloroplasts are green. What is crucial is the high reflectiveness of the near-infrared band which is invisible to the human eye. It is a result of the cellular structure and stoma on the leaf surface. Healthy plants show the highest reflectiveness of this spectrum band. If the plant is stressed, which is also demonstrated by a change in the porous structure, then the reflectiveness becomes lower, typically before the stress demonstrates itself in the visible part of the spectrum (e.g. by changing the leaf colour).
Another step is the use of a suitable multispectral camera, i.e. a sensor which is capable of separately capturing the above components of the light spectrum which are typical for plants. It makes it possible to compare individual isolated bands and use them for further calculations. The plant species, growth phase and various types of stress or infestation can be identified based on the course of the spectral curve. We use the Parrot Sequoia sensor which is equipped with a light sensor to ensure that measurements are not affected by a change in solar radiation and weather conditions during the flight. The data retrieved within the wavelength interval of approx. 1,300 - 3,000 nm (captured by a thermal camera), where the reflectiveness is primarily based on the water content, make it possible to analyse the water stress of plants and field irrigation systems.
Low reflectiveness of the visible component and, on the other hand, high reflectiveness of the infrared component is used for the establishment of "vegetation indexes" which are in most cases based on (in different variations) the proportion of the two spectrum components. These indexes are used to map vegetation and to identify and visualize quantitative indicators in the image area. NDVI and NDRE are the typical representatives, modifications include e.g. OSAVI index which minimizes the optical impact of soil, and EVI which uses the blue component to suppress atmospheric influences.
These indexes are used to predict crop yields, prepare application maps for agricultural variable dosing machinery, and to identify vegetation growth anomalies in a timely manner. On the other hand, images in the visible part of the spectrum are used to inspect the harvest and to quantify damage in the event of natural disasters, and thermal camera data help to optimize irrigation. However, the data always need to be analysed in the context of the given plant, time and place of imaging, soil properties and a number of other factors. Therefore we analyse the data in cooperation with the VÚRV (Crop Research Institute) in Prague.