Unmanned Aerial Vehicles Image Processing With the Use of a Neural Network

Abstract

Transport infrastructure facilities are critically important. To ensure their functioning, it is necessary to apply tracking methods that provide a high degree of protection. The article deals with the issues of unauthorized intrusion of foreign objects controlling, in order to prevent a dangerous impact on the infrastructure of high-speed transport. In this regard, it is proposed to conduct round-the-clock surveillance using unmanned aerial vehicles.
Due to the fact that the range of UAV’s action distance is limited, therefore, it is proposed to use a remote method of detecting the intrusion of objects on the infrastructure with the use of an optical cable OK. The joint use of UAVs and OK allows to create a reliable system that provides control over the intrusion on the infrastructure. Special video cameras (thermal imagers, Lidar) are installed on unmanned aerial vehicles, providing inspection of the invasion area during day and night time. Since video recording devices have different resolution, the task is to apply methods for integrating data with different resolution and processing them by a neural network. The implementation of infrastructure tracking systems requires increasing demands on the network structure. One of the tasks set in this article is the development of the structure of the intrusion detection network on the high–speed ground transport infrastructure.