Algorithms of a Multi-Position Radar System Using Neural Networks for Ornithological Safety Tasks

Detecting birds near airfields is one of the most urgent tasks of radars, as the number of aircraft collisions with birds is only increasing every year. In this article, we discuss the possibility of using neural networks to solve the problem of bird detection. The range of applications of neural network technologies in radar tasks is relevant, for example, in tasks of temporal and spectral filtering and object recognition. The main advantage of neural networks is that they can recognize deeper patterns in the data. In the learning process, neurons are able to respond to the information received in accordance with the principles of generalization, thereby solving the task assigned to them. In this article, the possibility of using neural networks as secondary signal processing will be analyzed for radar tasks. The paper analyzes two filtering algorithms: the first one is based on filtering the estimates of the state vector by Kalman, while the other algorithm uses an artificial neural network to predict the estimation of the state vector. For these secondary processing algorithms, an efficiency comparison based on location detection errors was performed. Experimental data obtained on the basis of a field experiment showed a slight gain in the accuracy of the neural algorithm over the Kalman filter and a loss in operational efficiency and computational costs.