Rostelecom a technique of the automated decodifying and comparison of circuits of arable lands

Main articles:

• Artificial intelligence in agriculture (HSC)
• IT in agro-industrial complex in the world
• IT in agro-industrial complex of Russia

2020: The announcement of a technique of the automated decodifying and comparison of circuits of arable lands

On October 9, 2020 the Rostelecom company announced development of a technique of the automated decodifying and comparison of circuits of arable lands. The provided technique is almost checked in pilot regions and showed efficiency and accuracy at the automated creation of circuits of arable lands.

According to the company, as pilot, polygons were selected from Egorlyksky district of the Rostov region and also Shimsky, Old Russian, Volotovsky and Soletsky areas of the Novgorod region.

Rostelecom developed a technique of comparison of circuits of arable lands using machine vision and neuronets

In borders of the selected administrative educations space pictures were received from satellites of remote sensing of Earth Sentinel-2 with the spatial resolution of 10 meters on pixel within the analyzed agricultural period. Then preprocessing of pictures with the necessary level of "noise" was made.

On the next stage, circuits of agricultural fields using algorithms of machine vision (Watershed) representing splitting the source image into a set of the areas covering it on the contrast gradient module were defined.

Arable areas classified using an author's algorithm using neural networks by a difference of values of indexes of vegetation (NDVI) and humidity (NDWI) between the spring and summer period for summer cultures, and the values of the index respectively shifted for winter crops.

The circuits of arable lands received with the help to this technique compared to the circuits provided by Analytical center of the Ministry of Agriculture of the Russian Federation. The developed technique showed efficiency and accuracy at the automated creation of circuits of arable lands in the territory of pilot administrative educations. Two types of discrepancies, discrepancies on borders of objects and discrepancies in object structure were revealed.

This technique of the automated decodifying and comparison of circuits of arable lands is developed in interaction with Analytical center of the Ministry of Agriculture of the Russian Federation.