Space technology to increase efficiency of crop cultivation, reduce fertiliser consumption and increase crop yields by about a quarter.
A group of scientists from Samara State Research University, Russia intend to offer farmers digital ‘vision’ systems for agricultural machinery based on technologies that were originally developed for space.
Space vision systems are traditionally used for searching for signs of life on Mars and investigation of other planets within our solar system. They are equipped with powerful optics and a set of sensors and are able to, among other things, analyse soil composition.
With a few innovations, these “space eyes” could be given to irrigation equipment, grain harvesters, and other agricultural equipment, scientists said.
The department of technical cybernetics of Samara University has designed a compact space hyperspectrometer to be installed on the Russian satellites and has been working on neural networks for the processing and classifying hyperspectral images of the Earth’s surface obtained from orbit.
The hyperspectral images show moisture and the mineral substances content in the soil, the presence of plant diseases, and even sources of pest spread, the press office said.
Text continues underneath image
“However, it turned out that the obtained hyperspectral information from a spacecraft cannot promptly meet the needs of precise agriculture, since its [gathering] takes a certain amount of time. Therefore, we turned to develop ground-based hyperspectral sensors,” explained Nikolai Kazansky, professor of the department of technical cybernetics at Samara University.
“The set of requirements for the ground-based sensors is significantly different from that for hyperspectral equipment for spacecraft. In the case of the space hyperspectrometers, the main thing is to obtain the maximum possible optical data, while for ground-based sensors, this is far from the primary task,” he added.
The scientists use sensor systems combining flat optics and high microrelief elements, which can perform several different tasks to obtain information about the state of soil and plants.
A hyperspectral camera turns into an extremely simple device, comparable in complexity to a conventional video camera
“For example, the combination of the phase functions of a harmonic lens and the phase function of a diffraction grating allows with only one element to form an image and to break it down into a spectrum. Thus, a hyperspectral camera turns into an extremely simple device, comparable in complexity to a conventional video camera, in which instead of a lens we use our optics, which simultaneously breaks down information into a spectrum and forms an image,” Kazansky said.
An operator could analyse the data, but in the future, this task should be performed by neural networks, which may also control agricultural machinery.
“We can install the hyperspectral sensors, for example, on the irrigation machine. After all, a hyperspectral image allows you to see many things that cannot be seen by a human eye in a conventional black-and-white or color image. And the sensor will instantly determine whether the field needs to be watered or not. We plan to use more than 50 spectral channels in the wavelength range of 0.4-1.05 microns,” Kazansky added.
Text continues underneath image
This technology is to make agriculture “extremely smart,” the scientists said. By installing the new system only on irrigation machines in Russia, agricultural producers will boost their yields by 25%, the scientists calculated. For instance, in 2019, Russia harvested 120.6 million tonnes of grain.
As part of the project, scientists promise to pay particular attention to the technical design of sensors – so they would be straightforward and cheap enough for mass use in agricultural machinery.
Hyperspectral sensors are expected to be installed not only on ground vehicles but also on drones, to assess the state of large areas of agricultural land immediately. The Samara State Agrarian University has already achieved preliminary agreements to try their technology on the local fields.
Under their project, the scientists received a grant for a four-year study. They plan to shape up the technology and introduce algorithms to reconstruct and analyse the hyperspectral images using deep learning methods of neural networks.