Skip to the main content

Original scientific paper

Application of neural networks for the detection and classification of artillery targets

Adriana Stubičar ; Ministry of Defence of the Republic of Croatia
Mario Šipoš ; Dr. Franjo Tuđman Croatian Defence Academy


Full text: english pdf 211 Kb

versions

page 41-58

downloads: 74

cite


Abstract

Neural networks have been in use since the 1950s and are increasingly prevalent in various domains of human activity, including military applications. Notably, GoogLeNet and convolutional neural networks, when appropriately trained, are instrumental in identifying and detecting individual objects within a complex set. In military scenarios, neural networks play a crucial role in the fire support process, especially when receiving target descriptions from forward observers. These networks are trained on image datasets to recognize specific features of individual elements or military objects, such as vehicles. As a result of this training, when presented with a new image, the network can accurately determine the type of vehicle, expediting the targeting process and enhancing the ability to provide a suitable response. This paper describes the application of neural networks for detecting and classifying artillery targets. It presents a specific problem and proposes a scientific solution, including explaining the methodology used and the results obtained.

Keywords

convolutional neural network, target detection, data augmentation

Hrčak ID:

312062

URI

https://hrcak.srce.hr/312062

Publication date:

22.12.2023.

Article data in other languages: croatian

Visits: 229 *