Analysis of Artificial Intelligence Methods for Detecting Drones Based on Radio Frequency Activity

The main illegal tasks of unmanned aerial vehicles are conducting reconnaissance, delivering prohibited goods, as well as striking transport and energy infrastructure facilities. The insignificant linear dimensions of drones reduce the effectiveness of their detection and their countering. Therefore, it is very important to know how to identify these devices, as well as be able to use them. Different data sets are needed to conduct such experiments. They are used for analysis to determine the most important elements. Therefore, data sets are a very valuable resource. This article describes the development of a classifier that determines the maximum information about unmanned aerial vehicles based on radio frequency data. The classifier is a convolutional neural network that can recognize the presence of a drone and the type of drone. The results showed that the developed product works better than existing analogues. Such indicators as accuracy, precision, recall and f1 metrics are 2% higher than those available for existing solutions.